AI Code Review Gate
Review an AI-generated code diff for the failure modes coding agents actually have — claimed-done-but-not-done, gamed or weakened tests, stubs passed off as complete, silent scope creep, hallucinated APIs, and security regressions. Returns an APPROVE or REQUEST CHANGES verdict with a completion check and severity-ranked fixes.