ralphy

name

ralphy

description

Autonomous AI coding orchestration using the Ralph Wiggum technique and Ralphy CLI. This skill should be used when running AI agents in continuous loops until tasks complete, orchestrating multi-task development with PRD/YAML files, configuring parallel agent execution, or implementing the Ralph Wiggum autonomous development methodology. Triggers on mentions of ralphy, ralph, autonomous loops, PRD-driven development, or multi-agent orchestration.

Ralphy & Ralph Wiggum Technique

Ralphy is an autonomous AI coding orchestrator that runs agents on tasks until done. Built on Geoffrey Huntley's Ralph Wiggum technique - a continuous loop methodology where AI agents iterate until completion.

Quick Reference

# Single task
ralphy "add dark mode"

# From PRD/YAML
ralphy --prd PRD.md
ralphy --yaml tasks.yaml

# Parallel execution
ralphy --parallel --max-parallel 5

# With branches and PRs
ralphy --branch-per-task --create-pr

Core Concepts

The Ralph Wiggum Technique

The foundational principle: run an AI agent in a loop until the job is done.

# Minimal form
while :; do cat PROMPT.md | claude ; done

Key insight: Failures inform improvements. When Ralph fails, tune the prompts like a guitar - add specific guidance based on observed failure patterns.

Three-Phase Architecture

1. Requirements Definition: Discuss project, identify Jobs to Be Done (JTBD), break into topics
2. Planning Loop: Gap analysis between specs and code, generate IMPLEMENTATION_PLAN.md