Create a Product Requirements Document (PRD)

Generate a comprehensive PRD that defines WHAT to build and WHY.

Arguments

$ARGUMENTS

• If --guided: Enable guided discovery (one question at a time)
• Otherwise: Batch 3-5 questions upfront

Task Tracking Setup

TASK_TRACKING = config_read("task_tracking.enabled", "false")

If TASK_TRACKING:
  PRD_WORKFLOW_ID = "prd-{timestamp}"
  PHASE_TASKS = {}

  PHASES = [
    {num: 1, name: "Clarifying Questions", active: "Gathering requirements"},
    {num: 2, name: "Generate PRD", active: "Generating PRD"},
    {num: 3, name: "Review & Options", active: "Reviewing PRD"},
    {num: 4, name: "Technical Expansion", active: "Expanding technical depth"},
    {num: 5, name: "Create Backlog Items", active: "Creating backlog items"}
  ]

  For each P in PHASES:
    PHASE_TASKS[P.num] = TaskCreate(
      subject: "Phase {P.num}: {P.name}",
      activeForm: P.active,
      metadata: {workflow: PRD_WORKFLOW_ID, phase: P.num}
    )

  # Sequential dependencies; phases 4 and 5 both depend on 3 (independent of each other)
  TaskUpdate(PHASE_TASKS[2], addBlockedBy: [PHASE_TASKS[1]])
  TaskUpdate(PHASE_TASKS[3], addBlockedBy: [PHASE_TASKS[2]])
  TaskUpdate(PHASE_TASKS[4], addBlockedBy: [PHASE_TASKS[3]])
  TaskUpdate(PHASE_TASKS[5], addBlockedBy: [PHASE_TASKS[3]])

---

Phase 1: Clarifying Questions

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[1], status: "in_progress")

Default Mode

Use AskUserQuestion to ask 3-5 essential questions in a single batch:

• Problem/Goal: What problem does this solve?
• Target Users: Who will use this?
• MVP Boundaries: What's the minimum to ship?
• Success Criteria: How will we measure success?
• Technical Context: Any existing systems to integrate?

Guided Mode (--guided)

Ask questions ONE AT A TIME:

1. Problem/Opportunity - What problem, who experiences it, how painful?
2. Solution Concept - How does your product solve this?
3. Target Users - Who specifically, technical level, current workaround?
4. Core Capabilities - 3-5 must-have features, what's NOT in v1?
5. Success Criteria - How will you know it's working?
6. Constraints - Technical, time, resource, dependencies?

After sufficient info, synthesize and confirm before PRD generation.

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[1], status: "completed")

---

Phase 2: Generate PRD

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[2], status: "in_progress")

Read the PRD template from assets/prd-template.txt (relative to this skill file).

Customize with user's answers:

• Fill in problem statement, user personas
• Write user stories with acceptance criteria
• Prioritize features using MoSCoW
• Set success metrics

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[2], status: "completed")

---

Phase 3: Review & Options

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[3], status: "in_progress")

1. Save to docs/prd/prd-[feature-name].md

2. Auto-Preview Check AUTO_PREVIEW = config_read("auto_preview", "false") If AUTO_PREVIEW is "true": Execute open docs/prd/prd-[feature-name].md

3. Present options via AskUserQuestion:

   • Done (Recommended) - Balanced PRD is sufficient
   • Expand with technical depth - Add API, Data Model, Security, Performance sections
   • Revise sections - Provide feedback on specific sections
   • Review and refine - Apply document-refinement skill, loop to Phase 3
   • Preview in editor (if not auto-previewed)

If "Revise sections": Loop back to Phase 3 without completing (keep in_progress).

If "Review and refine":

Apply document-refinement skill to docs/prd/prd-[feature-name].md
→ Auto-fix minor issues in PRD file
→ Present refinement report to user
→ Loop to Phase 3

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[3], status: "completed")

---

Phase 4: Technical Expansion (If Requested)

If user selected "Expand with technical depth": If TASK_TRACKING: TaskUpdate(PHASE_TASKS[4], status: "in_progress") Else: If TASK_TRACKING: TaskUpdate(PHASE_TASKS[4], status: "completed") Skip to Phase 5

Add sections from assets/technical-expansion.txt:

• API Specifications (endpoints, schemas, auth)
• Data Model with Mermaid ERD
• Security Considerations (AuthN/AuthZ, OWASP)
• Performance & Scalability (SLOs, scaling strategy)

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[4], status: "completed")

---

Phase 5: Create Backlog Items (Optional)

If user accepted backlog creation: If TASK_TRACKING: TaskUpdate(PHASE_TASKS[5], status: "in_progress") Else: If TASK_TRACKING: TaskUpdate(PHASE_TASKS[5], status: "completed") Skip to Output

Offer backlog creation:

• Create Must Have items - From Must Have features only
• Create all items - From all prioritized features
• Skip

If accepted:

1. Read backlog configuration from project CLAUDE.md
2. For each user story/feature:
   • Extract title and acceptance criteria
   • Set priority (Must=p1, Should=p2, Could=p3)
   • Create item using configured backend

If TASK_TRACKING: TaskUpdate(PHASE_TASKS[5], status: "completed")

Cleanup

If TASK_TRACKING:
  AUTO_CLEANUP = config_read("task_tracking.auto_cleanup", "true")
  If AUTO_CLEANUP:
    For each TASK in PHASE_TASKS.values():
      If TASK.status != "completed":
        TaskUpdate(TASK, status: "completed")

---

Output

Save to: docs/prd/prd-[feature-name].md

Create directory if needed.