ralph

You can enforce reviewer sign-off on PRD user stories by using a PRD-driven persistence loop that iterates through prd.json, verifying evidence for each story and requiring tiered critic approval before marking the task complete.

Preventing partial completion of multi-iteration code changes requires running evidence-based verification and acceptance criteria checks per user story, ensuring every story passes before the workflow concludes.

To maintain session persistence across retries for a hotfix workflow, use a stateful iteration loop that refines prd.json and tracks progress, ensuring mandatory review and regression re-verification upon resumption.

Yes, ralph supports selecting different reviewer profiles for task triage by using the --critic flag, allowing you to choose architect, critic, or codex-style approval for completion verification.

To run regression testing after deslop cleanup in a PRD workflow, perform an optional cleanup pass followed by a mandatory regression re-verification step to ensure acceptance criteria still pass.

You should avoid using a stateful iteration loop for incident triage when the task requires immediate deployment without mandatory acceptance criteria verification or reviewer approval.