What problem does it solve? Single-reviewer code reviews miss edge cases and suffer from individual blind spots, letting critical bugs reach production. This Skill runs two independent blind judge sub-agents in parallel against the same target, synthesizes their findings, and iterates fixes until both judges pass or the review escalates. ## Core Features & Use Cases - Parallel Blind Review: Launches two independent judge sub-agents simultaneously with identical criteria, ensuring no cross-contamination between reviewers. - Verdict Synthesis: Classifies findings as Confirmed, Suspect, or Contradiction, and distinguishes real warnings from theoretical ones to avoid nit-picking loops. - Fix and Re-judge Cycle: Delegates a separate Fix Agent for confirmed issues, then re-launches both judges, escalating to the user after two iterations if issues remain. - Project Standards Injection: Resolves project-specific skill registry rules and injects them into judge and fix prompts so reviews match codebase conventions. - Use Case: Before merging a significant feature, trigger the protocol to have two adversarial reviewers independently audit the code, apply confirmed fixes, and only approve once both judges return clean verdicts. ## Quick Start Ask the agent to run judgment day on the files or feature you want reviewed, for example by saying "run judgment day on the auth module".
