Ralph Converter

Convert PRDs to prd.json format for autonomous execution by the Ralph agent

The Ralph Converter skill (/ralph) converts markdown PRDs into the prd.json format that Ralph uses for autonomous execution. It ensures stories are properly sized, ordered by dependencies, and have verifiable acceptance criteria.

Overview

The skill takes a PRD (markdown file or text) and converts it to prd.json in your project’s ralph directory. The conversion process:

1. Validates story sizes (must fit in one context window)
2. Orders stories by dependencies (schema → backend → UI)
3. Ensures acceptance criteria are verifiable
4. Adds required criteria (typecheck, browser verification)
5. Archives previous runs if needed

Usage

Invoke the skill with any of these trigger phrases:

Convert tasks/prd-user-auth.md to prd.json
Turn this PRD into ralph format
Create prd.json from this
Generate ralph json for the notification system PRD

Or explicitly load it:

Load the ralph skill and convert [path-to-prd] to prd.json

Output Format

The skill generates a prd.json file with this structure:

prd.json

{
  "project": "[Project Name]",
  "branchName": "ralph/[feature-name-kebab-case]",
  "description": "[Feature description from PRD title/intro]",
  "userStories": [
    {
      "id": "US-001",
      "title": "[Story title]",
      "description": "As a [user], I want [feature] so that [benefit]",
      "acceptanceCriteria": [
        "Criterion 1",
        "Criterion 2",
        "Typecheck passes"
      ],
      "priority": 1,
      "passes": false,
      "notes": ""
    }
  ]
}

Field Descriptions

project · string · required

Project name (derived from PRD or repository)

branchName · string · required

Git branch name in kebab-case, prefixed with ralph/

description · string · required

Brief feature description from PRD introduction

userStories · array · required

Array of user story objects ordered by dependency and priority

userStories[].id · string · required

Sequential ID: US-001, US-002, etc.

userStories[].title · string · required

Short descriptive title from PRD

userStories[].description · string · required

User story in standard format: “As a [user], I want [feature] so that [benefit]”

userStories[].acceptanceCriteria · array · required

Verifiable criteria that define “done”. Always includes “Typecheck passes” and “Verify in browser using dev-browser skill” for UI stories.

userStories[].priority · number · required

Execution order (1 = highest). Based on dependencies, then document order.

userStories[].passes · boolean · required

Whether story is complete. Always starts as false.

userStories[].notes · string · required

Learnings from implementation. Always starts empty.

The Number One Rule: Story Size

Each story must be completable in ONE Ralph iteration (one context window).

Ralph spawns a fresh AI instance per iteration with no memory of previous work. If a story is too big, the LLM runs out of context before finishing and produces broken code.

Right-sized Stories ✅

• Add a database column and migration
• Add a UI component to an existing page
• Update a server action with new logic
• Add a filter dropdown to a list

Too Large ❌ (Split These)

• “Build the entire dashboard” → Split into: schema, queries, UI components, filters
• “Add authentication” → Split into: schema, middleware, login UI, session handling
• “Refactor the API” → Split into one story per endpoint or pattern

Rule of thumb: If you cannot describe the change in 2-3 sentences, it’s too big.

Story Ordering: Dependencies First

Stories execute in priority order. Earlier stories must not depend on later ones.

Correct Order ✅

1. Schema/Database Changes

   Database migrations, table creation, column additions

2. Server Actions / Backend Logic

   API endpoints, server functions, business logic

3. UI Components

   Frontend components that use the backend

4. Dashboard/Summary Views

   Higher-level views that aggregate data

Wrong Order ❌

1. US-001: UI component (depends on schema that doesn't exist yet)
2. US-002: Schema change (should have been first)

Acceptance Criteria: Must Be Verifiable

Each criterion must be something Ralph can CHECK, not something vague.

Good Criteria ✅

• “Add status column to tasks table with default ‘pending’”
• “Filter dropdown has options: All, Active, Completed”
• “Clicking delete shows confirmation dialog”
• “Typecheck passes”
• “Tests pass”

Bad Criteria ❌

• “Works correctly” (too vague)
• “User can do X easily” (not measurable)
• “Good UX” (subjective)
• “Handles edge cases” (which ones?)

Required Criteria

The skill automatically adds to every story:

"Typecheck passes"

For stories with UI changes, it also adds:

"Verify in browser using dev-browser skill"

Frontend stories are NOT complete until visually verified. Ralph will use the dev-browser skill to navigate to the page, interact with the UI, and confirm changes work.

Conversion Rules

The skill follows these rules when converting:

1. Each user story becomes one JSON entry - One-to-one mapping
2. IDs are sequential - US-001, US-002, etc.
3. Priority based on dependencies - Then document order
4. All stories start incomplete - passes: false, empty notes
5. branchName derived from feature - Kebab-case, prefixed with ralph/
6. Typecheck always added - Every story includes “Typecheck passes”

Splitting Large PRDs

If a PRD has big features, the skill splits them into smaller stories:

Original Story

### US-001: Add user notification system
**Description:** As a user, I want to receive notifications about important events.

**Acceptance Criteria:**
- [ ] Notifications stored in database
- [ ] Notification bell icon in header
- [ ] Dropdown panel shows notifications
- [ ] Can mark notifications as read
- [ ] Preferences page to control notifications

Split Into Smaller Stories

[
  {
    "id": "US-001",
    "title": "Add notifications table to database",
    "description": "As a developer, I need to store notifications in the database.",
    "acceptanceCriteria": [
      "Create notifications table with id, user_id, message, read, created_at",
      "Generate and run migration",
      "Typecheck passes"
    ],
    "priority": 1
  },
  {
    "id": "US-002",
    "title": "Create notification service",
    "description": "As a developer, I need a service to create and query notifications.",
    "acceptanceCriteria": [
      "createNotification() function saves to database",
      "getUnreadNotifications() returns unread for user",
      "markAsRead() updates read status",
      "Typecheck passes"
    ],
    "priority": 2
  },
  {
    "id": "US-003",
    "title": "Add notification bell icon to header",
    "description": "As a user, I want to see when I have notifications.",
    "acceptanceCriteria": [
      "Bell icon shows in header for authenticated users",
      "Badge shows unread count if > 0",
      "Clicking bell toggles dropdown panel",
      "Typecheck passes",
      "Verify in browser using dev-browser skill"
    ],
    "priority": 3
  },
  {
    "id": "US-004",
    "title": "Create notification dropdown panel",
    "description": "As a user, I want to see my notifications in a dropdown.",
    "acceptanceCriteria": [
      "Dropdown shows list of notifications (most recent first)",
      "Unread notifications visually distinct",
      "Empty state when no notifications",
      "Clicking notification marks as read",
      "Typecheck passes",
      "Verify in browser using dev-browser skill"
    ],
    "priority": 4
  },
  {
    "id": "US-005",
    "title": "Add notification preferences page",
    "description": "As a user, I want to control which notifications I receive.",
    "acceptanceCriteria": [
      "Preferences page accessible from dropdown",
      "Checkboxes for each notification type",
      "Saves preferences to database",
      "Typecheck passes",
      "Verify in browser using dev-browser skill"
    ],
    "priority": 5
  }
]

Each story is now one focused change that can be completed and verified independently.

Example Conversion

Here’s a complete example of converting a PRD:

Input PRD

tasks/prd-task-status.md

# Task Status Feature

Add ability to mark tasks with different statuses.

## Requirements
- Toggle between pending/in-progress/done on task list
- Filter list by status
- Show status badge on each task
- Persist status in database

Output JSON

prd.json

{
  "project": "TaskApp",
  "branchName": "ralph/task-status",
  "description": "Task Status Feature - Track task progress with status indicators",
  "userStories": [
    {
      "id": "US-001",
      "title": "Add status field to tasks table",
      "description": "As a developer, I need to store task status in the database.",
      "acceptanceCriteria": [
        "Add status column: 'pending' | 'in_progress' | 'done' (default 'pending')",
        "Generate and run migration successfully",
        "Typecheck passes"
      ],
      "priority": 1,
      "passes": false,
      "notes": ""
    },
    {
      "id": "US-002",
      "title": "Display status badge on task cards",
      "description": "As a user, I want to see task status at a glance.",
      "acceptanceCriteria": [
        "Each task card shows colored status badge",
        "Badge colors: gray=pending, blue=in_progress, green=done",
        "Typecheck passes",
        "Verify in browser using dev-browser skill"
      ],
      "priority": 2,
      "passes": false,
      "notes": ""
    },
    {
      "id": "US-003",
      "title": "Add status toggle to task list rows",
      "description": "As a user, I want to change task status directly from the list.",
      "acceptanceCriteria": [
        "Each row has status dropdown or toggle",
        "Changing status saves immediately",
        "UI updates without page refresh",
        "Typecheck passes",
        "Verify in browser using dev-browser skill"
      ],
      "priority": 3,
      "passes": false,
      "notes": ""
    },
    {
      "id": "US-004",
      "title": "Filter tasks by status",
      "description": "As a user, I want to filter the list to see only certain statuses.",
      "acceptanceCriteria": [
        "Filter dropdown: All | Pending | In Progress | Done",
        "Filter persists in URL params",
        "Typecheck passes",
        "Verify in browser using dev-browser skill"
      ],
      "priority": 4,
      "passes": false,
      "notes": ""
    }
  ]
}

Archiving Previous Runs

Before writing a new prd.json, the skill checks if there’s an existing one from a different feature:

1. Read current prd.json (if it exists)
2. Check if branchName differs from the new feature’s branch name
3. If different AND progress.txt has content:
   • Create archive folder: archive/YYYY-MM-DD-feature-name/
   • Copy current prd.json and progress.txt to archive
   • Reset progress.txt with fresh header

The ralph.sh script handles archiving automatically when you run it, but if you’re manually updating prd.json between runs, the skill will archive for you.

Workflow

1. Have a PRD Ready

   Create a PRD using the PRD Generator skill or write one manually in tasks/prd-[feature-name].md

2. Convert to JSON

   Invoke the Ralph converter:

   Load the ralph skill and convert tasks/prd-feature-name.md to prd.json
   
3. Review and Edit

   Review the generated prd.json. Check:

   • Story sizes are appropriate
   • Dependencies are in correct order
   • Acceptance criteria are verifiable
4. Run Ralph

   Start the autonomous execution loop:

   ./scripts/ralph/ralph.sh
   

Validation Checklist

Before saving prd.json, the skill verifies:

• Previous run archived (if prd.json exists with different branchName)
• Each story is completable in one iteration (small enough)
• Stories are ordered by dependency (schema → backend → UI)
• Every story has “Typecheck passes” as criterion
• UI stories have “Verify in browser using dev-browser skill” as criterion
• Acceptance criteria are verifiable (not vague)
• No story depends on a later story

Next Steps

After generating prd.json:

1. Review - Check story sizes and ordering
2. Commit - Add the prd.json to version control
3. Run Ralph - Execute ./scripts/ralph/ralph.sh to start autonomous implementation
4. Monitor - Check progress.txt for learnings and git log for commits

Common Issues

Stories Too Large

If Ralph consistently fails to complete stories:

• Each story should be 2-3 sentences maximum
• Split into smaller, more focused stories
• Check for multiple logical changes in one story

Wrong Dependency Order

If stories fail due to missing dependencies:

• Ensure schema changes come before backend logic
• Backend logic must exist before UI can use it
• Move dependent stories to later in the priority order

Vague Acceptance Criteria

If Ralph can’t determine when a story is complete:

• Replace “works correctly” with specific behaviors
• Add concrete UI states or API responses
• Include specific file paths or function names