ralph

Convert PRD documents into prd.json user stories for the Ralph autonomous agent system.

Updated May 19, 2026
One-click install
npx skills add https://github.com/azusagasaku/--claude-config --skill ralph-azusagasaku
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ralph
Source: https://github.com/azusagasaku/--claude-config/tree/main/skills/ralph
Command: npx skills add https://github.com/azusagasaku/--claude-config --skill ralph-azusagasaku

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? The Ralph autonomous agent system requires a specific prd.json format with right-sized, dependency-ordered user stories, and manually converting a markdown PRD into that structure is error-prone and often produces stories too large for a single agent iteration. ## Core Features & Use Cases - PRD to prd.json Conversion: Transforms markdown PRDs into Ralph's JSON schema with project, branchName, description, and userStories fields. - Story Sizing and Ordering: Splits oversized features into stories completable in one iteration and orders them by dependency (schema, backend, UI). - Verifiable Acceptance Criteria: Enforces checkable criteria like "Typecheck passes" and browser verification for UI stories. - Use Case: You have a PRD describing a notification system. Use this Skill to split it into six focused stories (database table, service, bell icon, dropdown, mark-as-read, preferences) and output a ready-to-run prd.json. ## Quick Start Convert this PRD into Ralph's prd.json format with properly sized and ordered user stories.

Frequently Asked Questions about ralph

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

FAQPage Schema
How do I convert a PRD to prd.json for Ralph?▼

Provide your markdown PRD and the Skill maps each requirement to a userStories entry with sequential IDs, priorities, and acceptance criteria. It derives the branchName from the feature name in kebab-case with a ralph/ prefix and writes the result to prd.json.

How big should a user story be for autonomous agents?▼

Each story must be completable in one Ralph iteration, meaning one context window. If you cannot describe the change in 2-3 sentences, split it into smaller stories such as schema changes, backend logic, and individual UI components.

What acceptance criteria should AI agent tasks have?▼

Criteria must be verifiable, such as "Add status column with default 'pending'" or "Typecheck passes", not vague statements like "works correctly". UI stories should also include browser verification using the dev-browser skill.

How are user stories ordered in prd.json?▼

Stories execute in priority order, so earlier stories must not depend on later ones. The correct order is schema and database migrations first, then backend logic, then UI components, and finally aggregate dashboard views.

What happens to an existing prd.json when starting a new feature?▼

If the existing prd.json has a different branchName and progress.txt has content, the Skill archives both into an archive/YYYY-MM-DD-feature-name/ folder and resets progress.txt before writing the new file.