ralph

Enforce PRD-driven iteration loops with reviewer verification for task completion.

Updated Mar 17, 2026
One-click install
npx skills add https://github.com/Rheinmir/skills-kit --skill ralph-rheinmir
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ralph
Source: https://github.com/Rheinmir/skills-kit/tree/main/skills/ralph
Command: npx skills add https://github.com/Rheinmir/skills-kit --skill ralph-rheinmir

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Ralph provides a self-sustaining loop that relentlessly pursues task completion until every user story passes verification, preventing premature or unchecked delivery.

Core Features & Use Cases

  • PRD-driven iteration: Breaks work into verifiable user stories tracked in prd.json, progressing only when criteria pass.
  • Persistent, multi-iteration execution: Maintains state across retries and iterations for long-running tasks.
  • Reviewer-verification gate: Requires fresh sign-off before marking completion to ensure quality.

Quick Start

Start Ralph to begin a PRD-driven iteration; it scaffolds prd.json if needed and guides you through story-by-story verification.

Frequently Asked Questions about ralph

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

FAQPage Schema
How do I enforce verifiable task completion for complex task management?

Verifiable task completion is enforced by a PRD-driven iteration loop that tracks user stories in prd.json, persisting state across multiple iterations until all acceptance criteria pass reviewer verification.

What is the best way to prevent premature task delivery in automated workflows?

To prevent premature task delivery, a reviewer-verification gate requires fresh sign-off before marking completion, ensuring mandatory deslop and regression checks are passed for every user story.

How does state persistence work during long-running task iterations?

State persistence during long-running task iterations is maintained by saving progress continuously to structured files like prd.json and progress.txt, allowing execution to resume cleanly across retries.

Do I need a PRD to start automated task verification?

A PRD is needed to start automated task verification, but the system scaffolds prd.json automatically if one is absent, guiding you through defining explicit acceptance criteria and story-by-story verification.

Can I use a structured iteration loop for multi-iteration task management?

A structured iteration loop supports multi-iteration task management by breaking work into verifiable user stories, holding execution in a self-sustaining cycle that relentlessly pursues completion until criteria pass.

Why does task verification require a reviewer gate before completion?

Task verification requires a reviewer gate to guarantee quality by preventing unchecked delivery, enforcing explicit deslop and regression checks, and requiring fresh sign-off before any user story is marked complete.