ck:code-review

Adversarial code review is a red-team approach to identifying hidden defects and security holes by actively probing edge cases and attempting to break changes. It enforces quality through evidence-led findings rather than surface-level inspection.

To run an adversarial code review, provide a clear diff or target code changes using input modes like pending changes, PR numbers, commit hashes, or full codebase scans. The pipeline then executes scout, review, adversarial, and verify stages.

Yes, you can scan a full codebase for security flaws using the codebase scan input mode. The adversarial reviewer performs an exhaustive assessment to uncover hidden defects and false assumptions across the provided code.

Yes, the review pipeline requires a clear diff, testable evidence, and verification before claiming completion. This evidence-driven approach ensures that all identified flaws are substantiated rather than speculative.

The best way to enforce code quality on pending changes is applying an adversarial review pipeline that actively attempts to break the modifications. This multi-stage workflow probes failure modes and requires verification before approval.