claude-config-audit
Audit the .claude configuration for valid frontmatter, discoverable agents and skills, missing skill references, stale slash-command plugin names, broad permissions, and hook/settings regressions.
SKILL.md.claude/skills/claude-config-audit/SKILL.md
GitHubClaude Config Audit
Use this skill before merging any change under .claude/**, CLAUDE.md, or agent-memory files.
Workflow
- Parse YAML frontmatter in every
.claude/**/*.mdfile. - Build inventories for agents, skills, commands, rules, imports, output styles, hooks, and memory.
- Verify every agent
skills:entry maps to.claude/skills/<name>/SKILL.md. - Search slash commands for stale
agent-skills:*references and browser-only workflows that do not apply to this terminal simulator. - Check orchestrators for overly broad delegation and confirm
/shiprequires DegenExus verification evidence. - Confirm
.claude/settings.jsonexists and contains conservative secret/destructive-command denies plus hook wiring. - Report findings as
BLOCKER,IMPORTANT, orSUGGESTIONwith exact file paths.
Required Local Commands
python .claude/hooks/validate-claude-config.py
python -m compileall -q .claude/hooks
Exit Criteria
- Frontmatter parses cleanly.
- Agent skill references are resolvable.
- Commands reference local agents/skills or documented project workflows only.
- Hooks and settings are present and auditable.