cross-modal-review

Cross-modal review captures your code diffs and spawns a different model to audit results, returning a verdict with evidence. It validates work against the originating skill's Contract to ensure output quality before finalization.

Cross-modal review handles model refusals using refusal routing, silently switching to the next model in the chain. This ensures your quality-assurance workflow continues without manual intervention when a review model refuses a task.

You need cross-modal review when validating produced artifacts like brain-page generation, analysis, or code diffs against an originating skill's Contract. It provides an independent validation layer by routing work through a different model.

Yes, cross-modal review implements a modular /codex handoff specifically for diff-review use cases. This allows you to route code changes directly to the codex tool for independent validation.

Cross-modal review reports findings with clear agreement and disagreement signals. It returns a verdict with evidence, comparing your work product against the originating skill's Contract and returning independent validation results.

Cross-modal review depends on the originating skill's Contract to validate results, meaning it requires a defined Contract to function. If a model refuses, it silently switches to the next in chain, but output quality is bounded by the available models.