Open-source tool uses adversarial AI reviewer to catch code plan flaws

A developer built agent-plan-review-loop, an open-source multi-agent system where an adversarial Claude reviewer reads the actual codebase before approving implementation plans. The reviewer defaults to rejecting plans and must be convinced they are sound, catching errors like non-existent file paths and missing edge cases. The system uses complexity-aware model routing to balance cost and thoroughness across different task types.