ralph

PRD-driven persistence loops enforce task completion by breaking work into prioritized user stories from prd.json, iterating until every story passes reviewer verification. This prevents complex AI work from ending in incomplete states by continuously rechecking acceptance criteria across retries.

Test-driven workflow verification is automated by wrapping parallel execution with session persistence and automatic retry-on-failure behavior. Progress is recorded in progress.txt while the system iterates until defined acceptance criteria are explicitly met and verified.

Agent orchestration applies to multi-iteration engineering work including code changes that require tests and build verification. It wraps execution with session persistence, automatically retrying failed tasks until all PRD user stories pass reviewer acceptance.

PRD execution includes mandatory post-review cleanup and clean session cancellation as part of its persistence loop. It also handles regression re-verification to ensure previously passing stories remain valid after new iterations.

AI-assisted tasks often finish incomplete because they lack PRD-based story tracking and fresh reviewer verification before declaring completion. Enforcing a PRD-driven persistence loop iterates until all user stories pass and are reviewer-verified, eliminating gaps.

Track multi-iteration progress using session persistence and progress.txt recording. The system maintains state across retries, wrapping parallel execution so that failed attempts automatically retry until all acceptance criteria are satisfied and verified.