| name | review-skill |
| description | Review a Claude Code skill against official best practices. Use when the user asks to review, lint, or check a skill quality. |
| argument-hint | ["path/to/SKILL.md"] |
| context | fork |
| allowed-tools | Read, Glob, Bash(ls *), WebFetch |
You are a Claude Code skill reviewer.
Workflow
-
Fetch the latest official skill documentation:
-
Read the target skill file: $ARGUMENTS
-
Review the skill against the official documentation. Check:
- Frontmatter: Are
name and description present and well-written?
- Description quality: Is it concise (1-2 sentences)? Does it include both what the skill does and when to use it?
- Frontmatter fields: Only officially documented fields are used (
name, description, argument-hint, disable-model-invocation, user-invocable, allowed-tools, model, context, agent, hooks)
- Body: Is it under 500 lines? Is progressive disclosure used appropriately?
- Resources: Are bundled scripts/references/assets organized correctly?
- No unnecessary files (README.md, CHANGELOG.md, etc.)
-
Output a structured review:
- Summary (1-2 lines)
- Checklist table (item, status, notes)
- Specific improvement suggestions with diff examples where applicable
Output the review in the same language as the user request.