How to Win
3 rounds. 3 agents per round. Same angles every round. Each round goes deeper.
Research skill that figures out what winning means for a specific situation. Deploys 3 research agents (Codebase Archaeologist, Domain Scout, Adversary) across 3 rounds. Each round reads previous findings and asks: "what did you miss?"
Model requirement: Use model: "opus" when dispatching researcher agents.
The Loop
ROUND 1: Deploy 3 researchers (Archaeologist, Scout, Adversary)
→
[Description truncada. Veja o README completo no GitHub.]