ralph

PRD-driven iteration prevents premature task completion by dividing complex tasks into discrete user stories with mandatory reviewer verification. It uses persistent execution and automatic retries to ensure every story passes acceptance criteria before being marked as done.

PRD-driven iteration tracks task progress across discrete user stories using a prd.json scaffold. It applies parallel, deterministic execution with mandatory reviewer verification, persisting state in progress.txt until all acceptance criteria pass.

To set up PRD-driven task iteration, provide a task description to start the loop, then supply a refined prd.json scaffold and concrete acceptance criteria for each user story. The system automatically tracks per-story progress until all stories pass verification.

Yes, PRD-driven task management supports parallel, persistent task execution with automatic retries. It enforces mandatory reviewer verification before completion, applying consistent deslop cleanups throughout the iteration process until all user stories pass.

Use PRD-driven iteration when complex tasks require persistent, multi-iteration work to meet strict acceptance criteria. It is ideal for scenarios where preventing premature "done" states is critical and every user story must pass reviewer verification.