cross-modal-review

Multi-model code review catches AI hallucinations by spawning secondary AI models to perform independent reviews against defined contractual requirements. This mandatory second-opinion quality gate verifies work products objectively before finalizing critical work.

Adversarial security testing for code diffs identifies potential injection vectors, race conditions, and data leaks before committing changes. You need this security-focused review to ensure high-integrity output and prevent vulnerabilities when finalizing critical work.

Contract-based grading validates code by evaluating work against the originating skill's specific promises rather than subjective quality metrics. This approach ensures the output meets defined contractual requirements through an objective, multi-model consensus verification process.

Model routing bypasses AI refusals during automated code reviews by automatically switching to a secondary model if the primary reviewer refuses a request. This refusal-aware routing ensures continuous validation without manual intervention or workflow disruption.

You need git-diff context to run an independent multi-model security check because it provides the necessary code changes for objective review. The validation process requires this integration with model-routing conventions to ensure high-integrity output and accurate adversarial testing.