ralph

To guarantee task completion, you can use a PRD-driven persistence loop that continues iterating with parallel agent execution until every user story in prd.json passes reviewer verification. This enforces mandatory verification before final delivery.

Automatic retry on failure in PRD-driven workflows works by wrapping parallel execution with session persistence, maintaining structured story tracking across retries. The loop continues iterations until all user stories pass the designated reviewer's verification.

Yes, you can configure a specific reviewer to verify user stories during iteration by selecting a reviewer via the --critic flag. This enforces mandatory verification before the workflow can mark the task as complete.

The best way to track progress across persistent retries is using structured story tracking built into the PRD scaffolding. It monitors the status of each user story in prd.json, ensuring automatic retries only target incomplete or unverified items.

Yes, PRD-driven iteration supports deslop passes to ensure end-to-end quality. The persistence loop enforces PRD scaffolding, progress tracking, and deslop passes before allowing the reviewer to verify and complete the user stories.

You should not use an automated persistence loop for agent orchestration if your task lacks a structured PRD or clear user stories. The loop strictly enforces PRD scaffolding and mandatory verification, which requires defined criteria to iterate against.