Ralph Skill

Convert PRD markdown documents to Ralph's JSON format for autonomous execution.

When to Use

Use this skill after creating a PRD with the prd skill. The user will say something like: "Load the ralph skill and convert tasks/prd-[name].md to prd.json"

Workflow

1. Read the PRD

bash

cat tasks/prd-[feature-name].md

2. Generate prd.json

Create scripts/ralph/prd.json with this structure:

json

{
  "featureName": "Feature Name from PRD",
  "branchName": "feature/kebab-case-name",
  "description": "Brief description from PRD overview",
  "userStories": [
    {
      "id": "US-001",
      "title": "Story title",
      "description": "What needs to be implemented",
      "acceptanceCriteria": [
        "Criterion 1",
        "Criterion 2",
        "Criterion 3"
      ],
      "priority": 1,
      "passes": false
    }
  ]
}

3. Story Mapping Rules

For each user story in the PRD:

PRD Field	JSON Field
Story ID (US-001)	`id`
Story title	`title`
User story text + technical notes	`description`
Acceptance criteria	`acceptanceCriteria` array
Order in document	`priority` (1 = first)
Always	`passes: false`

4. Branch Naming

Generate a clean branch name:

•Prefix with feature/
•Use kebab-case
•Keep it short but descriptive
•
Examples:
- •feature/user-authentication
- •feature/dashboard-analytics
- •feature/trip-details-modal

5. Save the File

bash

mkdir -p scripts/ralph
cat > scripts/ralph/prd.json << 'EOF'
{
  "featureName": "...",
  ...
}
EOF

6. Initialize Progress File

bash

cat > scripts/ralph/progress.txt << 'EOF'
# Ralph Progress Log
# Learnings and context for future iterations

## Feature: [Feature Name]
## Started: [Date]
## Branch: [branch-name]

---

EOF

7. Copy Prompt Template (if not exists)

Check if scripts/ralph/prompt.md exists. If not, the user needs to copy it from the Ralph installation.

8. Confirm with User

code

✅ Created scripts/ralph/prd.json with [N] stories
✅ Initialized scripts/ralph/progress.txt

To start Ralph:
  ./scripts/ralph/ralph.sh [max_iterations]

To check status:
  ./scripts/ralph/ralph-status.sh

Stories will be implemented in order:
1. US-001: [title]
2. US-002: [title]
...

Quality Checks

Before finalizing, verify:

•Story sizes - Each should be ~30 min of focused work
•Dependencies - Lower priority numbers should not depend on higher
•Acceptance criteria - Clear, testable, 3-5 per story
•Branch name - Valid git branch name

If stories seem too large, suggest splitting them.

Example Conversion

Input (PRD excerpt):

markdown

### US-001: Add trips database table
**As a** developer
**I want to** have a trips table in the database
**So that** I can store trip information

**Acceptance Criteria:**
- [ ] Migration creates trips table
- [ ] Table has id, name, start_date, end_date columns
- [ ] Migration runs without errors

**Technical Notes:**
- Use Drizzle ORM migration
- Follow existing schema patterns in db/schema.ts

Output (prd.json excerpt):

json

{
  "id": "US-001",
  "title": "Add trips database table",
  "description": "Create database migration for trips table with id, name, start_date, end_date columns. Use Drizzle ORM migration following patterns in db/schema.ts.",
  "acceptanceCriteria": [
    "Migration creates trips table",
    "Table has id, name, start_date, end_date columns",
    "Migration runs without errors"
  ],
  "priority": 1,
  "passes": false
}

Tips

•Keep descriptions concise but include technical hints
•Acceptance criteria should be checkable, not vague
•Frontend stories should include "Verify in browser" as acceptance criteria
•Test stories should specify what tests to write, not just "add tests"