Audit AI Agent Instructions

Workflow

1. Inspect agent instruction files and identify scope, precedence, and intended tools.
2. Remove duplicate, contradictory, stale, or unsafe instructions.
3. Add safety rules for secrets, destructive commands, dirty worktrees, version checks, and verification.
4. Keep instructions procedural and concise.
5. Validate links and examples.

Verification

• Instructions are clear, non-conflicting, and safe for real repositories.