fix(web): match skill chip icons to file chips and add sent-message tooltips

[Mnigos]

Two small skill chip fixes that came out of testing workspace skill discovery. The skill icon now renders at the same optical size as file-chip icons (the package glyph fills its viewBox edge to edge, file-type icons carry intrinsic padding). And sent-message chips get the description tooltip the composer already had, with the InlineSkill prop type carrying the description fields through MessagesTimeline and ChatMarkdown, which previously narrowed them away.

Also documents the workspace-scoped skill behavior that #8778 and #9210 shipped in docs/user/composer.md, since neither added docs.

Earlier revisions of this PR carried the workspace discovery itself; that landed piecewise through #8778 by @UtkarshUsername and #9210 by @anirudhsama (based on the approach from #4031 by @D3OXY), so this PR is down to the chip polish. Chip rendering has 7 focused tests.

Demo

Screen.Recording.2026-09-02.at.00.57.59.mov

Screenshots

Implemented with Claude Code (Claude Fable 5).

---

Note

Medium Risk
New RPC spawns Codex app-server processes per workspace lookup (mitigated by timeout and cache); cwd resolution touches orchestration projection data but stays read-only with ownership checks.

Overview
Fixes the $ skill picker and sent-message $name chips using skills from the open project/worktree instead of the server’s startup directory.

Adds provider.listWorkspaceSkills (read-scoped RPC) with contracts for input/result/source: workspace | snapshot. The server resolves cwd from thread.worktreePath ?? project.workspaceRoot, validates thread ownership, and calls an optional listSkillsForCwd on provider instances—with a 30s TTL cache per (instanceId, cwd) that does not mutate provider snapshots. Codex spawns a short-lived app-server in the workspace and calls skills/list; Claude uses discoverClaudeSkills for the resolved cwd; other drivers return source: "snapshot" so clients keep snapshot skills.

On the web, useWorkspaceSkills fetches the workspace catalog for the composer ($ menu, slash skills) and the timeline so chips survive sending. Skill chip UI is aligned (shared wrapper class, smaller skill icon) and timeline chips get description tooltips like the composer.

^{Reviewed by Cursor Bugbot for commit 26e4a23f392d6c6c72e285dbb3fd001aee61177a. Configure here.}

Note

Align skill chip icons with file chips and add description tooltips to sent messages

• Introduces shared inline-chip wrapper and dedicated skill-icon classes in composerInlineChip.ts so skill chips in both the composer and chat timeline render with consistent 1em icon sizing
• Expands the shared InlineSkill type to include description and shortDescription, and adds resolveInlineSkillDescription which trims and prefers a non-empty short description before falling back to the full description
• Conditionally wraps skill chips in a top-positioned tooltip when a description resolves; chips without a description render without a tooltip trigger
• Updates ChatMarkdown, MessagesTimeline, CollapsibleUserMessageBody, and UserMessageBody to consume the shared InlineSkill type instead of local ServerProviderSkill projections
• Adds server-rendering tests in SkillInlineText.test.tsx covering known/unknown tokens, punctuation adjacency, shared wrapper usage, and tooltip behavior

^{Macroscope summarized e1679bc.}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 26e4a23f392d6c6c72e285dbb3fd001aee61177a. Configure here.}

[macroscopeapp]

Approvability

Verdict: Approved at e1679bc

Macroscope's review found this PR approvable — This PR is limited to web UI polish for existing skill chips, adding conditional description tooltips, aligning icon sizing, and carrying existing metadata through the renderer. The change is self-contained, backward-compatible, and supported by focused tests, with no schema, infrastructure, security, billing, or default-setting impact.

^{You can add or adjust custom eligibility rules. Learn more.}

Dismissing prior approval to re-evaluate 01313dd

Dismissing prior approval to re-evaluate dab755e

Dismissing prior approval to re-evaluate e1679bc