audit-agent-readiness

Use when judging whether agents can work in a repo safely and alone, or when setup offers it. Mines the repo's history for mistakes agents repeat, probes a disposable worktree, reads agent instructions for prose that works around a code defect, and returns improvements grouped by the highest layer where each fix works. Trigger on intent like "is this repo ready for agents", "audit agent readiness", or "why do agents keep tripping here".