self-review-on-stop

Automatically audit an AI response by triggering a Stop hook to review the final text for contradictions, factual errors, and format drift. If issues are found, it returns a corrected response; otherwise, it replies with a single space.

Silent self-review is an automated process that verifies coherence, factuality, and formatting consistency across your final text outputs. It runs during top-level Stop turns to ensure response quality without requiring manual intervention.

To set up a stop-hook for response quality, trigger the audit workflow after delivering a top-level response. The workflow verifies coherence and replies with a single space if the output is clean or a fully corrected response otherwise.

Yes, when the self-review workflow finds unsupported claims, contradictions, or formatting issues, it replies with the full corrected response prefixed with a specific marker. If the text is clean, it only returns a single space.

Yes, you can trigger a self-review audit on demand through explicit self-review requests. The dedicated audit workflow will then execute to verify response coherence, factuality, and format consistency.

A limitation of using a stop-hook for format consistency checking is that it operates silently, returning a single space if clean or a fully corrected response otherwise, which may overwrite outputs without explicit user confirmation.