Skip to content
Get packs

Ralph Persistence Loop

Keeps an agent iterating on a task via a PRD of user stories until each is reviewer-verified

Source Download skill
Repository
Yeachan-Heo/oh-my-claudecode
Stars
40,512
Licence
MIT
Files
1 file

What it does

Once installed, when you ask the agent to keep going until a complex task is truly done, it works through a structured list of user stories (a prd.json file) one at a time, re-running its work across iterations and checking each story against acceptance criteria before marking it complete. It also runs feedback commands such as build, test and lint via a verify step, and can use a separate reviewer (architect, critic, or codex) to sign off before finishing. The skill file is truncated in the provided text, so some details of the loop are not fully visible.

Great for

  • Developers with multi-step coding tasks that tend to end up partially finished
  • Teams who want acceptance criteria and reviewer sign-off before a task is declared done
  • Users of the oh-my-claudecode toolkit who want a persistence loop distinct from autopilot

Skip it if

  • Quick one-shot fixes, where the loop adds overhead; the file itself suggests delegating directly to an executor agent
  • Exploratory work or planning before committing, which the file says should use the plan skill instead
  • Environments where unattended runs could modify code without review, since the skill relies on a headless mode (omc ralph afk) and repeated iterations that edit a repository

Try asking

  • “ralph: finish migrating the auth module to the new token format and keep going until all tests pass”
  • “ralph --critic=critic add pagination to the orders API with reviewer sign-off”
  • “ralph --no-deslop refactor the billing service; skip the cleanup pass”
  • “keep going until done: fix every failing lint rule in this repo”
Step-by-step guide: how to use Ralph Persistence LoopInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    ralph.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip ralph.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip ralph.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give a clear task description with testable acceptance criteria so the PRD stories are meaningful
  • Check the generated prd.json and feedback commands before a long run, since the loop executes repository commands
  • Use --critic to choose a reviewer that fits the task; architect is the default

What's inside

  • SKILL.md

Redistributed unmodified from Yeachan-Heo/oh-my-claudecode under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.