Adversarial Code Review + Security/Test Bundle

A pre-commit review skill for AI coding agents that works in three passes. First, an adversarial reviewer actively tries to break the changed code: hunting for logic errors, bad edge-case handling, concurrency bugs, and swallowed errors. Second, a security checklist pass covers injection, auth and authorization flaws, hardcoded secrets, unsafe deserialization, and path traversal. Third, the agent writes missing unit tests for the changed code in your project's own test framework. Output is a severity-graded report (blocker / major / minor) using an included rubric, plus generated tests. Runs against a git diff or a file list before every commit or PR. Built for solo developers, small dev shops, and agencies that ship client code and want a second, deliberately hostile pair of eyes on every change.