agent-review

@crsmithdev

Review all AI-runtime config — CLAUDE.md, hooks, skills, agent personas — against `src/rules/agent/RULES.md` and its four sub-surface rule files. Scans, presents findings grouped by severity, asks at the approval gate, applies approved fixes, verifies. Covers four sub-surfaces in a single walk: config (CLAUDE.md, settings.json), hooks (src/core/hooks/*.ts), skills (src/skills/*/SKILL.md, skill-rules.json), personas (src/agents/*.md). Cross-sub-surface drift (persona referencing a renamed skill, hook writer→consumer pairs, skill-rules entries vs SKILL.md files on disk) is a first-class finding. Triggers on /agent-review, /audit agent, "audit my config", "audit the agent setup", "audit my hooks", "are my hooks wired", "audit my skills", "find orphaned skills", "audit my agents", "find agent drift", "check cross-domain drift", "what's broken in my setup".