| name | reviewing-skills |
| description | Use when reviewing or improving Claude Code, Codex, or compatible agent skills. Checks SKILL.md frontmatter, trigger boundary, progressive disclosure, no-op instructions, duplication, sediment, sprawl, examples, and validation readiness. Trigger for skill review, SKILL.md review, skill quality check, trigger debugging, or improving an existing skill. |
Reviewing Skills
Review skills for predictable agent behavior, not just readability.
Workflow
-
Identify the target skill directory and read SKILL.md completely.
-
Read references/best-practices.md before producing findings.
-
Check frontmatter name and description for trigger precision and exclusions.
-
Check the body for clear steps, completion criteria, progressive disclosure, duplication, no-ops, sediment, and sprawl.
-
Inspect bundled references, scripts, assets, and evals when present.
-
Return prioritized findings before optional rewrites.
Report Shape
Use: Critical, Warning, Info, and Suggested rewrite when a concrete replacement is useful.
Detailed Reference
For non-trivial implementation, review, or refactoring work, read references/details.md before giving final guidance. It contains the detailed rules, examples, smells, and migration notes that do not belong in the short invocation body.
Output
Lead with actionable findings and file references. If asked to fix, make the smallest changes that improve invocation, execution, and validation.