Ralph Planning Skill

Generate a Product Requirements Document (PRD) with prioritized user stories from requirements. This prepares a project for autonomous iterative development with /ralph-execute.

Invocation

/ralph-plan <requirements> [options]

Arguments:

• <requirements> - Feature description (inline text, file path, or URL)
• --branch NAME - Branch name (default: auto-generated with ralph/ prefix)
• --max-stories N - Maximum stories to generate (default: 10)

Examples:

/ralph-plan "Build a user authentication system with login and registration"
/ralph-plan ./requirements.md --branch ralph/auth-system
/ralph-plan --max-stories 15

Execution Flow

Phase 1: Gather Requirements

Step 1.1: Parse Input

If <requirements> is:

• File path: Read the file contents
• URL: Fetch and extract requirements
• Inline text: Use directly
• Not provided: Ask the user what feature they want to build

Step 1.2: Analyze Codebase Context

Before generating stories, understand the existing codebase:

# Identify project type and structure
ls -la
cat package.json 2>/dev/null

Look for:

• Existing patterns in the codebase
• UI component library in use
• State management approach
• Testing patterns
• Routing solution

Step 1.3: Identify Quality Gate Commands

Read package.json scripts section and determine available commands:

Phase 2: Ask Clarifying Questions

CRITICAL: Before generating stories, ask 3-5 clarifying questions with lettered options.

Use the AskUserQuestion tool to get clarity on:

1. Scope and approach decisions
2. UI/UX preferences
3. Technical constraints
4. Priority trade-offs

Example Questions:

Before I generate stories, I need to clarify a few things:

1. Authentication approach:
   A) Email/password only
   B) OAuth (Google, GitHub)
   C) Magic links
   D) All of the above

2. UI framework preference:
   A) Use existing component library
   B) Build from scratch
   C) Mix of both

3. Error handling:
   A) Toast notifications
   B) Inline error messages
   C) Both

Please respond like "1A, 2A, 3C" with your choices.

Phase 3: Generate PRD

Step 3.1: Create User Stories

Transform requirements into atomic, testable user stories following these rules:

Story Sizing Rules:

• ✅ Add database column with default value
• ✅ Create single UI component
• ✅ Add one API endpoint
• ✅ Add form validation
• ❌ Build entire dashboard (too big - split into schema, queries, components, filters)
• ❌ Complete authentication system (too big - split into login, register, reset, session)

Each story MUST fit in one context window (~5-15 minutes implementation)

Story Template:

{
  "id": "US-001",
  "title": "Brief action-oriented title (max 80 chars)",
  "description": "As a [user], I want [capability] so that [benefit]. Implementation should [specific guidance].",
  "acceptanceCriteria": [
    "Component renders without errors",
    "TypeScript compiles with no errors",
    "Follows existing patterns in src/components/",
    "Verify in browser using visual-polish-inspector"  // For UI stories
  ],
  "priority": 1,
  "status": "pending",
  "failureCount": 0,
  "requiresBrowserVerification": true  // true for UI stories
}

Step 3.2: Order Stories by Dependency

Stories MUST be ordered so earlier ones don't depend on later ones:

1. Foundation - Types, schemas, context providers, utilities
2. Backend - API routes, server actions, database changes
3. Components - UI components that depend on backend
4. Pages/Views - Pages that compose components
5. Integration - Final wiring, polish, edge cases

Step 3.3: Write prd.json

Create .ralph/ directory and write the PRD:

mkdir -p .ralph

Write to .ralph/prd.json:

{
  "projectName": "[Extracted from requirements]",
  "branchName": "ralph/[feature-slug]",
  "createdAt": "[ISO timestamp]",
  "updatedAt": "[ISO timestamp]",
  "qualityGates": {
    "typecheck": "npm run check",
    "lint": "npm run lint",
    "test": null,
    "build": null
  },
  "config": {
    "maxIterations": 10,
    "maxFailuresPerStory": 3,
    "autoCommit": true
  },
  "userStories": [...],
  "summary": {
    "totalStories": N,
    "passed": 0,
    "blocked": 0,
    "pending": N,
    "iterationsRun": 0
  }
}

Phase 4: Initialize Tracking

Step 4.1: Initialize progress.txt

Write header to .ralph/progress.txt:

================================================================================
RALPH PROGRESS LOG - [Project Name]
Branch: [branch-name]
Started: [ISO timestamp]
================================================================================

--- CODEBASE PATTERNS (Consolidated) ---
// This section is updated after each iteration
// Reusable patterns also go to AGENTS.md

[Initial patterns discovered during analysis]

================================================================================

Step 4.2: Create/Switch to Feature Branch

git checkout -b [branch-name] 2>/dev/null || git checkout [branch-name]

Phase 5: Present Summary

Display the generated PRD summary:

RALPH PRD GENERATED
━━━━━━━━━━━━━━━━━━━

Project: [Name]
Branch: [branch-name]
Stories: [N] user stories

Priority Order:
1. US-001: [Title]
2. US-002: [Title]
...

Quality Gates:
✓ TypeScript: npm run check
✓ Lint: npm run lint
○ Test: not configured
○ Build: disabled

Files Created:
- .ralph/prd.json
- .ralph/progress.txt

Next Step:
Run `/ralph-execute` to begin autonomous development.
Run `/ralph-execute --dry-run` to preview the first story.

Error Handling

Quality Checklist

Before completing, verify:

• Asked 3-5 clarifying questions
• Stories are small enough (1-3 files each)
• Stories are ordered by dependency
• Each story has testable acceptance criteria
• UI stories include browser verification requirement
• Quality gates are configured correctly
• Feature branch created/checked out

Anti-Patterns to Avoid

1. Vague acceptance criteria - "Works correctly" is NOT acceptable
2. Oversized stories - If it touches >3 files, split it
3. Missing dependencies - Story 5 can't depend on Story 10
4. Skipping questions - Always ask clarifying questions
5. Generic descriptions - Be specific about implementation approach

Output Files

After successful execution:

1. .ralph/prd.json - Complete PRD with stories
2. .ralph/progress.txt - Initialized log with codebase analysis
3. Git branch created/checked out