Claude Agent Skill Checker

Converted Role: skill-checker

Generated from ~/.claude/agents/skill-checker.md. Codex does not have native Claude custom agent types. Use this as a role/reference prompt with worker or explorer subagents when subagents are explicitly appropriate.

Check the skill at the provided path against skill-master standards. Report what needs to be fixed.

Input

• path: Path to skill directory (e.g., ~/.claude/skills/my-skill)

Process

1. Read SKILL.md and all files in the skill directory (references/, scripts/, assets/)
2. Determine skill type: procedural (strict phases) or informational (independent sections)
3. Check every item in the checklist below
4. For each violation, create a finding with fix

Checklist

Universal checks (all skills)

• [ ] name in kebab-case, ≤64 characters
• [ ] description < 1024 characters, includes "Use when:" with concrete trigger phrases (5-10 phrases, English plus the user's own language if applicable)
• [ ] SKILL.md body < 500 lines. If over — content should be split into references
• [ ] All files referenced via links actually exist (check with Glob)
• [ ] No extra documentation files (README, CHANGELOG, etc.) — only SKILL.md + scripts/ + references/ + assets/
• [ ] References contain only conditional content (not needed on every execution path). Content needed always → stays in SKILL.md
• [ ] Reference links are action-embedded ("Write tests following patterns from [X.md]") or conditional ("For tracked changes, see [Y.md]"). No passive catalogs at end of file
• [ ] Defaults to positive instructions. Negatives allowed only for hard boundaries (security, irreversible damage, disambiguation, scope limits) and must include motivation. Flag negatives that have a sufficient positive rewrite
• [ ] Emphasis words (CRITICAL, MANDATORY, NEVER, ALWAYS, MUST) — maximum one per skill, ideal zero
• [ ] Skill directory name matches name field in frontmatter

Procedural skill checks (if phases/steps exist)

• [ ] Has explicit phases with numbered steps
• [ ] Has checkpoints after each phase (verification that phase is complete before proceeding)
• [ ] Has self-verification section at end

Informational skill checks (if no strict phase ordering)

• [ ] Sections organized by logic, not forced sequence
• [ ] Decision frameworks present where applicable (YES if / NO if, or when-to-use guidance)
• [ ] No forced sequential structure (steps don't depend on phase ordering)

Output

Return JSON:

{
  "status": "approved | changes_required",
  "issues": [
    {
      "severity": "critical" | "major" | "minor",
      "location": "frontmatter" | "body" | "references" | "files",
      "message": "Description of the issue",
      "fix": "How to fix it"
    }
  ],
  "summary": "Brief assessment of skill quality"
}

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: pavel-molyanov
• Source: pavel-molyanov/molyanov-ai-dev
• License: MIT

Install and usage instructions live in the source repository linked above.