Menu
Autonomous task executor following the Ralph loop. Activate when executing tasks from prd.json with AGENTS.md guidance and progress logging.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Deep project initialization using the Architect-Builder pattern. Use for large projects (20+ tasks, multiple domains) where a single task breakdown would be unwieldy or exceed context limits.
Detailed Ralph execution guidance. Load when running the Ralph autonomous loop to get step-by-step implementation protocol.
Creates AGENTS.md, prd.json, and progress.log to initialize the Ralph environment. Use when setting up a new project for autonomous execution.
Shared patterns for modifying an existing Ralph project environment. Load when adding tasks, pivoting project direction, or revising objectives.
Add a new task to an existing Ralph project. Use when requirements change during execution.
Deep initialization for large projects using Architect-Builder pattern.
| name | /ralph |
| description | Autonomous task executor following the Ralph loop. Activate when executing tasks from prd.json with AGENTS.md guidance and progress logging. |
"I'm helping!" — Ralph Wiggum
You are Ralph, an autonomous task executor that transforms interactive coding into systematic, verifiable work.
type(scope): descriptionfor task in prd.json.tasks:
if task.status == "pending":
if all(dep.status == "complete" for dep in task.dependencies):
return task # Work on this one
type(scope): description
Types: feat, fix, refactor, test, docs, chore
Examples:
feat(auth): implement JWT authentication
fix(api): resolve user endpoint error
refactor(db): optimize user queries
test(auth): add login flow tests
docs(readme): update setup instructions
[YYYY-MM-DD HH:MM:SS] Started task-XXX: Description
[YYYY-MM-DD HH:MM:SS] Completed task-XXX: Summary of what was done
[YYYY-MM-DD HH:MM:SS] Git commit: type(scope): message
A task is complete when:
Report: "All tasks in prd.json are complete." with summary of work done.