Iterate through PRD stories until all pass and a reviewer verifies completion.
npx skills add https://github.com/dropsyoon/oh-my-claudecode --skill ralph-dropsyoon
Please help me install this Agent Skill. Skill: ralph Source: https://github.com/dropsyoon/oh-my-claudecode/tree/main/skills/ralph Command: npx skills add https://github.com/dropsyoon/oh-my-claudecode --skill ralph-dropsyoon
Ralph prevents complex work from being marked “done” prematurely by continuing until every PRD user story is fully satisfied and verified, then performing a final reviewer sign-off (with an optional cleanup pass).
passes: true and reviewer verification are achieved.prd.json, then replaces generic acceptance criteria with task-specific, verifiable ones.progress.txt, persists state across iterations, and reruns only what’s needed.architect, critic, or codex) to validate against the exact acceptance criteria.ai-slop-cleaner on changed files and re-checks tests/build/lint before completing.Tell your agent: "ralph, keep going until done and verify each PRD story until every requirement passes."
High-intent search queries and answers about installing and using this skill.
PRD-driven persistence loops prevent premature completion by executing multi-iteration tasks until all acceptance criteria pass and a designated reviewer verifies the final state.
To generate task-specific acceptance criteria, the system automatically creates or loads a prd.json file and replaces generic requirements with verifiable, task-specific conditions to guide the execution loop.
Yes, the verification loop is designed for multi-iteration engineering tasks needing reliable convergence across retries, tracking progress via a structured file and persisting state to rerun only what is needed.
Yes, you can use tiered reviewer verification by selecting an architect, critic, or codex reviewer to validate the implemented code against the exact acceptance criteria defined in the prd.json.
The best way to clean up AI-generated code is running an automated cleanup tool on changed files, followed by mandatory post-cleanup regression re-verification to ensure tests, builds, and lint checks still pass.
ArvorCoGenerate a structured PRD with goals, user stories, and acceptance criteria.
INNERJOINTAutomate PRD-driven execution loops until all user stories pass verification.
Goblin1024Identify and remove AI-generated slop in changed files using a regression-tests-first cleanup workflow.
d-o-hubOrchestrate workflow agents in iterative refinement loops with configurable termination modes.