ralph_skill

Overview

This skill converts existing PRDs (markdown or plain text) into Ralph's prd.json format for autonomous execution. It produces a validated, dependency-ordered set of small user stories with required acceptance criteria and metadata ready for the Ralph workflow. Use it when you need a runnable prd.json that follows Ralph's story sizing and ordering rules.

How this skill works

Provide a PRD or paste the markdown/text. The converter extracts project name, feature title/description, and breaks the PRD into discrete user stories sized to complete in one Ralph iteration. It assigns sequential IDs, sets priorities by dependency order, appends required criteria like "Typecheck passes", and outputs a prd.json with branchName in ralph/kebab-case format.

When to use it

• You have an existing PRD and need prd.json for Ralph
• Preparing a feature for automated implementation with Claude Code
• Splitting large features into iteration-sized stories
• Archiving or replacing an existing prd.json before a new run
• Ensuring acceptance criteria are verifiable and ordered by dependency

Best practices

• Keep each story completable in one Ralph iteration (2–3 sentence scope)
• Order stories: schema → backend → UI → dashboards
• Always include "Typecheck passes" in every story's acceptanceCriteria
• Add "Verify in browser using dev-browser skill" for UI stories
• Archive existing prd.json if branchName differs before saving new file

Example use cases

• Convert a Task Status PRD into prd.json with separate stories for schema, UI badge, toggle, and filter
• Split a large notification system PRD into multiple focused stories: schema, service, UI components, preferences
• Generate a prd.json from a bugfix-focused PRD that only needs a migration and a server action
• Prepare an app feature PRD for autonomous builds by ensuring verifiable acceptance criteria