1# Harness Skills Capabilities Reference
 2
 3Source of truth for what each AI coding harness supports in terms of agent skills.
 4Used to inform provider configs in `scripts/lib/transformers/providers.js`.
 5
 6Last verified: 2026-03-24
 7
 8## Official Documentation
 9
10| Harness | Docs URL |
11|---------|----------|
12| Claude Code | https://code.claude.com/docs/en/skills |
13| Cursor | https://cursor.com/docs/context/skills |
14| Gemini CLI | https://geminicli.com/docs/cli/skills/ |
15| Codex CLI | https://developers.openai.com/codex/skills |
16| GitHub Copilot (Agents) | https://code.visualstudio.com/docs/copilot/customization/agent-skills |
17| Kiro | https://kiro.dev/docs/skills/ |
18| OpenCode | https://opencode.ai/docs/skills/ |
19| Pi | https://github.com/badlogic/pi-mono/blob/main/packages/coding-agent/docs/skills.md |
20| Trae | TBD (no official skills docs found yet) |
21| Rovo Dev | https://support.atlassian.com/rovo/docs/extend-rovo-dev-cli-with-agent-skills |
22
23## Spec Compliance
24
25All harnesses follow the [Agent Skills specification](https://agentskills.io/specification) to varying degrees. The spec defines these frontmatter fields: `name`, `description`, `license`, `compatibility`, `metadata`, `allowed-tools`.
26
27Provider-specific extensions beyond the spec: `user-invocable`, `argument-hint`, `disable-model-invocation`, `allowed-tools` (extended syntax), `model`, `effort`, `context`, `agent`, `hooks`, `subtask`, `mcp`.
28
29## Frontmatter Support
30
31Fields marked with * are spec-standard. Others are provider extensions.
32
33| Field | Claude Code | Cursor | Gemini | Codex | Copilot | Kiro | OpenCode | Pi | Rovo Dev |
34|-------|:-----------:|:------:|:------:|:-----:|:-------:|:----:|:--------:|:--:|:--------:|
35| `name`* | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
36| `description`* | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
37| `license`* | Yes | Yes | Ignored | No | Yes | Yes | Yes | Yes | Yes |
38| `compatibility`* | Yes | Yes | Ignored | No | Yes | Yes | Yes | Yes | Yes |
39| `metadata`* | Yes | Yes | Ignored | No | Yes | Yes | Yes | Yes | Yes |
40| `allowed-tools`* | Yes | No | Ignored | No | No | No | Yes | Yes | Yes |
41| `user-invocable` | Yes | No | No | No | Yes | No | Yes | No | Yes |
42| `argument-hint` | Yes | No | No | No | Yes | No | Yes | No | Yes |
43| `disable-model-invocation` | Yes | Yes | No | No | Yes | No | Yes | Yes | TBD |
44| `model` | Yes | No | No | No | No | No | Yes | No | No |
45| `effort` | Yes | No | No | No | No | No | No | No | No |
46| `context` | Yes | No | No | No | No | No | No | No | No |
47| `agent` | Yes | No | No | No | No | No | Yes | No | No |
48| `hooks` | Yes | No | No | No | No | No | No | No | No |
49
50Notes:
51- Gemini CLI validates only `name` and `description`; other spec fields are parsed but ignored.
52- Codex CLI uses a separate `agents/openai.yaml` sidecar for extended metadata (icons, branding, MCP tools, invocation control).
53- Kiro recognizes `user-invocable` and `disable-model-invocation` per community reports but does not formally document them.
54- Unknown fields are silently ignored by all harnesses.
55
56## Skill Directory Structure
57
58| Harness | Native directory | Also reads |
59|---------|-----------------|------------|
60| Claude Code | `.claude/skills/` | - |
61| Cursor | `.cursor/skills/` | `.agents/skills/`, `.claude/skills/`, `.codex/skills/` |
62| Gemini CLI | `.gemini/skills/` | `.agents/skills/` |
63| Codex CLI | `.agents/skills/` (primary) | - |
64| GitHub Copilot | `.github/skills/` | `.agents/skills/`, `.claude/skills/` |
65| Kiro | `.kiro/skills/` | - |
66| OpenCode | `.opencode/skills/` | `.agents/skills/`, `.claude/skills/` |
67| Pi | `.pi/skills/` | `.agents/skills/` |
68| Trae China | `.trae-cn/skills/` | TBD |
69| Trae International | `.trae/skills/` | TBD |
70| Rovo Dev | `.rovodev/skills/` | `~/.rovodev/skills/` (user-level) |
71
72All harnesses support the `{skill-name}/SKILL.md` directory structure with optional `reference/`, `scripts/`, and `assets/` subdirectories.
73
74## Placeholder / Variable Substitution
75
76Claude Code supports runtime variable substitution directly in SKILL.md bodies: `$ARGUMENTS`, `$0`-`$N`, `${CLAUDE_SKILL_DIR}`, `${CLAUDE_SESSION_ID}`. No other harness supports substitution in skills.
77
78Some harnesses have separate "custom commands" systems (distinct from skills) with their own substitution:
79
80| Harness | Command system | Substitution syntax |
81|---------|---------------|-------------------|
82| Gemini CLI | `.gemini/commands/` (TOML) | `{{args}}`, `!{shell}`, `@{file}` |
83| Codex CLI | `.codex/prompts/` | `$ARGNAME` |
84| OpenCode | `.opencode/commands/` | `$ARGUMENTS`, `$1`-`$N`, `` !`shell` `` |
85
86Our build system handles cross-provider placeholders at compile time via `replacePlaceholders()` for `{{model}}`, `{{config_file}}`, `{{ask_instruction}}`, and `{{available_commands}}`.