* Follow-up of #2592

We've been testing this for 2 weeks, and it's working great.
This is a very useful enhancement to reduce token usage, cost, and make
smaller models behave better.

This is particularly significant for local models. These models are tiny,
and with a long enough system prompt they have little context window left
for work.

It's still possible to have the old behavior by setting:

    export CRUSH_SHORT_TOOL_DESCRIPTIONS=0

Some models like Kimi K2.5 and GLM 5 simply don't handle well that many
content. When a big page or JSON API is fetch, it can break Crush
sessions.

We're drastically reducing the truncation size here, to 1/10 of what we
had before. It may look too much at first look, but 100KB it's actually
very reasonable. For example, the entire Task guide, which is
significantly large, is only 61KB:

https://taskfile.dev/docs/guide.md

* fix(lsp): mitigate stale diagnostics

* fix(lsp): log warning when workspace change notification fails

💘 Generated with Crush

Assisted-by: GLM-5.1 via Crush <crush@charm.land>

* test(lsp): add unit tests for diagnostics settling behavior

💘 Generated with Crush

Assisted-by: GLM-5.1 via Crush <crush@charm.land>

* refactor(lsp): remove blocking diagnostic wait from post-edit notifications

💘 Generated with Crush

Assisted-by: GLM-5.1 via Crush <crush@charm.land>

* chore: add async diagnostic waits

* Update internal/agent/tools/diagnostics.go

Co-authored-by: Andrey Nering <andreynering@users.noreply.github.com>

* fix(lsp): stop timer leaks in WaitForDiagnostics

Replace time.After with time.NewTimer + defer Stop() to prevent
unreclaimable timer allocations on early returns.

💘 Generated with Crush

Assisted-by: GLM-5.1 via Crush <crush@charm.land>

---------

Co-authored-by: Andrey Nering <andreynering@users.noreply.github.com>

For example, the GitHub MCP server can fetch diffs. This makes those
diffs render as such, rather than just showing their raw output.

Descriptions should describe when a skill applies, not how it does its
job. Leaving language like "using go doc", "the jq command-line
processor", etc. in the trigger encourages the model to guess its way
through without reading SKILL.md.

Models frequently skipped viewing a skill's SKILL.md and inferred its
behavior from the description alone; making the load mandatory in
<critical_rules> prevents LLMs from taking that shortcut.

For example: Apple Terminal.

Do not run these two new tests in parallel.

💘 Generated with Crush

Assisted-by: Claude Opus 4.7 via Crush <crush@charm.land>

Also, two local skills now exist for Crush development:

* builtin-skills, for adding new builtin skills
* shell-builtins, for adding new shell builtins

- Display loaded/errored skills in sidebar and landing page
- Use pubsub event to carry skill states directly to UI, avoiding
  global mutable state
- Subscribe to skills discovery events via broker, consistent with
  MCP/LSP patterns
- Rename State to DiscoveryState for clarity

It takes a lot of time to re-record these VCRs for so many models.
Sometimes they fail intermittently, and you need to manually re-record
again and again until it works.

Let's keep a single provider and models which should work well enough for
us.

When a session is interrupted mid-tool-call (Ctrl-C, network timeout),
the assistant message contains a tool_use block but no corresponding
tool_result is ever recorded. On session resume, the Anthropic API
rejects the conversation with invalid_request_error, permanently
locking the session.

The existing filterOrphanedToolResults handles the inverse case
(tool_result without a matching tool_use). This commit adds the
symmetric fix: during preparePrompt, detect tool_use IDs that have
no matching tool_result and inject a synthetic error tool_result so
the API accepts the history and the session can continue.

Co-authored-by: Christian Rocha <christian@rocha.is>
Co-authored-by: Andrey Nering <andreynering@users.noreply.github.com>

See: https://github.com/charmbracelet/ultraviolet/pull/109