cross-modal-review

A cross-modal review provides a second opinion on code changes before merge by sending the work product and its originating contract to a different AI model. It forces a structured quality gate that verifies the output fulfills what was promised before finalizing anything.

Cross-modal contract compliance review verifies a work product satisfies the originating skill’s promised Contract. It uses model routing with refusal-chain fallback to ensure reliable evaluation and reports structured PASS or ISSUES grading with specific evidence citations.

Refusal routing handles model failures during diff review by silently switching to the next model in the review chain if one refuses. It escalates to the user only if all available models refuse to perform the evaluation.

AI code review with strict user-sovereignty cannot automatically apply suggestions to your codebase. The reviewer provides transparent grading with specific citations and agreement signals, but you retain full control to decide what changes to implement.

You should use an adversarial challenge for security-sensitive updates when performing a code-review handoff. This structured mode provides a deeper independent diff review to ensure significant security changes meet contract compliance before merging.