My agents wrote over 200 bad "watch pr" scripts over the last few months. If you have a "babysit" skill for watching PRs, send the following prompt to your agent: I want you to audit my history with Claude Code, Codex and other agents on this machine. Look for every time I asked for a PR to be watched or babysat. 1. How many times was the "babysitting" logic re-invented? 2. How many of these implementations had flaws you can see? 3. Roughly how many tokens (and how much $ at API prices) was…