cross-modal-review

To validate code diffs against predefined contracts, this skill orchestrates a secondary AI model review process, enforcing a mandatory second-opinion check before finalizing critical work to mitigate hallucinations and logic errors.

Multi-model peer review for quality assurance is a validation technique that evaluates work products like code diffs and architectural decisions by routing them through a secondary AI model to ensure alignment with predefined contracts.

Refusal routing handles model failures during security-focused code review by implementing silent model-switching, automatically falling back to a secondary model if the primary reviewer refuses a request or encounters an issue.

Yes, you can use adversarial testing to identify security vulnerabilities and injection vectors in architectural decisions by enforcing an adversarial challenge review process before finalizing complex refactors or new skill definitions.

Automated code review via this skill evaluates work based on contract-based grading, measuring work products against the originating skill's specific promises rather than relying on subjective quality assessments.

The limitation of automated model switching for code validation is that the system provides structured feedback for user-sovereign decision making rather than auto-fixing issues, meaning the user must still manually resolve identified injection vectors or contract misalignments.