angry-ralph

Converts a feature spec into reviewed, tested codebase via six-phase pipeline with resume capability.

5|Updated Feb 26, 2026
One-click install
npx skills add https://github.com/Custos/angry-ralph --skill angry-ralph
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: angry-ralph
Source: https://github.com/Custos/angry-ralph/tree/main/skills/angry-ralph
Command: npx skills add https://github.com/Custos/angry-ralph --skill angry-ralph

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

angry-ralph is a unified orchestration pipeline that converts a user-provided feature spec into a fully implemented, tested codebase through six sequential phases, combining decomposition, planning, adversarial review, and TDD-driven execution.

Core Features & Use Cases

  • End-to-end planning: decompose specs, plan implementations, and coordinate multi-LLM adversarial review to surface issues early.
  • Deterministic execution with gating: drive a Ralph Loop that enforces TDD red-green cycles and gating on tests.
  • Resumable sessions: persist state and artifacts so sessions can resume after interruptions.
  • Adversarial quality control: leverage external reviewers to improve plans and code before delivery.
  • Use Case: Plan and implement a feature from a high-level spec and obtain a reviewed, tested codebase ready for integration.

Quick Start

Load a feature spec and start the six-phase pipeline from decomposition through final review.

Frequently Asked Questions about angry-ralph

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I orchestrate a feature spec into a tested codebase with TDD?

The pipeline uses a six-phase process that decomposes your feature spec, creates a detailed plan, applies adversarial multi-LLM review, and drives TDD red-green cycles to produce reviewed, tested code.

What is adversarial multi-LLM review in a code generation pipeline?

Adversarial multi-LLM review is a structured quality control gate where external LLMs critique implementation plans and code to surface issues early, ensuring robust output before final integration.

How do I resume an interrupted TDD pipeline session?

You resume an interrupted TDD pipeline using built-in state persistence that saves artifacts and progress, allowing the orchestration pipeline to recover and continue execution seamlessly after interruptions.

Can I use adversarial review gates for large feature decomposition?

Yes, adversarial review gates support large feature decomposition by splitting detailed plans into sections and enforcing structured review gates across each section before final integration.

What are the limitations of using a structured TDD pipeline for feature implementation?

The pipeline enforces strong prerequisites and structured review gates, meaning it requires a complete feature spec upfront and may halt execution if strict TDD gating tests fail during the Ralph Loop cycles.