ralph

To ensure PRD user stories pass acceptance criteria without silent failures, use an iterative execution loop that continuously runs tasks until every story is verified by a reviewer. This approach replaces generic criteria with task-specific, verifiable checks before implementation.

The best way to track task execution progress across multiple iterations is through session-persistent progress tracking stored in a prd.json file. This method breaks tasks into individual user stories and maintains state across retries until all completion criteria are met.

Regression testing after a deslop cleanup pass works by performing a mandatory re-verification of all previously approved user stories. After initial reviewer sign-off, the cleanup pass removes code bloat, followed by a regression guardrail to ensure the cleanup process breaks nothing.

Yes, you need a PRD scaffold to initialize task execution for multi-iteration engineering work. The process requires PRD scaffold initialization and refinement to break down tasks into user stories with verifiable acceptance criteria before iterative implementation begins.

Automated story refinement replaces generic acceptance criteria to prevent unverifiable partial progress. By generating task-specific, verifiable checks from the scaffold before implementation, the system ensures that iterative execution targets precise completion conditions rather than vague requirements.

Yes, you can use session persistence to prevent stopping at partial progress during regression testing. By storing user stories in a persistent prd.json file, the execution loop continues across retries until all acceptance criteria pass and receive reviewer sign-off.