Runs a structured audit of your Claude Code and Codex setup to catch instruction drift, missing verifiers, failing MCP servers, and AI maintainability problems before they rot your codebase. It tiers projects by size (simple/standard/complex) and only runs deep inspections when you ask for them or when the summary pass finds critical ambiguity. The collector script checks agent config across AGENTS.md and runtime files, tests MCP tools live, and flags hotspots where AI-written code is getting hard to maintain. Findings come with copy-paste fixes, not vague suggestions. Useful when Claude starts ignoring instructions, when validation commands go stale, or when you suspect your agent setup has drifted from reality.
npx skills add https://github.com/tw93/waza --skill health