Prd

PRD Skill

Create concise, high-quality PRDs that align teams on what to build and why. A PRD is an alignment tool, not a bureaucratic artifact. Every section should earn its place by reducing ambiguity or enabling better decisions.

Important: This skill produces PRDs. Do NOT start implementing.

The Workflow

Step 1: Clarify

Before writing, use the AskUserQuestion tool to gather essential information where the user's prompt is ambiguous. Ask 1-4 focused questions with 2-4 options each. The tool automatically adds an "Other" option for custom input.

Focus questions on:

• Problem/Goal: What problem does this solve? Why now?
• Target users: Who is this for?
• Core functionality: What are the key actions?
• Scope/Boundaries: What should it NOT do?
• Success criteria: How do we know it worked?

Example questions:

Question: "What is the primary goal of this feature?"
Options:
- "Improve onboarding experience" - Make it easier for new users to get started
- "Increase retention" - Keep existing users engaged
- "Reduce support burden" - Decrease common support requests

Question: "Who is the target user?"
Options:
- "New users only" - Optimized for first-time experience
- "All users" - Benefits everyone
- "Power users/admins" - Advanced functionality for experienced users

When to ask questions:

• The user's prompt lacks key context (problem, users, scope, or success criteria)
• You need to choose between multiple valid interpretations
• Architectural decisions require user preference (e.g., scope, priorities, tradeoffs)

Skip this step if:

• The user provides comprehensive detail upfront
• The user explicitly asks to skip clarification
• You're reviewing an existing PRD (not creating a new one)

Step 2: Generate the PRD

Use the structure below. Adapt section depth to the size of the effort — a small feature gets a 1-2 page PRD, a new product gets 3-5 pages. Cut sections that don't apply rather than filling them with filler.

Step 3: Save

• Format: Markdown (.md)
• Location: tasks/prd-[feature-name].md (kebab-case)
• If a different output format or location is requested, use that instead.

Core Principles

These guide every PRD this skill produces:

1. Concise over comprehensive — A PRD that gets read beats one that covers everything.
2. Problems over solutions — Lead with user pain and business context, not implementation.
3. Evidence over assertions — "Users struggle with X" is weak; "43% of support tickets cite X" is strong.
4. Living over locked — Mark incomplete sections [WIP] rather than filling them with platitudes.
5. Exciting over exhaustive — The best PRDs make the team want to build this.

PRD Structure

1. Header

2. Problem & Opportunity

The most important section — it justifies the entire effort.

• What problem are you solving? For whom? Why now?
• Back it with evidence: support tickets, usage data, user interviews, competitive pressure
• What happens if you do nothing?

Don't describe the feature here — describe the world without it.

3. Goals & Success Metrics

Specific, measurable objectives. Not aspirations.

• Primary metric: The single number that tells you this worked
• Secondary metrics: Supporting indicators
• Guardrail metrics: Things that must NOT get worse

Bad: "Improve engagement." Good: "Increase weekly active usage by 15% within 90 days of launch."

4. Target Users

• Primary persona (optimized for when tradeoffs arise)
• Secondary personas (supported but not prioritized)

5. User Stories

Each story should be small enough to implement in one focused session.

Format:

### US-001: [Short Title]
**Description:** As a [user], I want [feature] so that [benefit].

**Acceptance Criteria:**
- [ ] Specific, verifiable criterion
- [ ] Another criterion

Acceptance criteria must be testable. "Works correctly" is bad. "Button shows confirmation dialog before deleting" is good.

6. Functional Requirements

Numbered list of specific capabilities:

- FR-1: The system must allow users to [action]
- FR-2: When a user clicks X, the system must [response]

Be explicit and unambiguous. Write for the implementer — they may be a junior developer or an AI agent. Avoid jargon or explain it. Use concrete examples where helpful.

7. Non-Goals (Out of Scope)

What this feature will NOT include and why. This prevents scope creep and manages expectations. Also note future considerations — things intentionally deferred but worth designing for.

8. Dependencies & Risks

Be honest about what could go wrong. Include both technical and business risks.

9. Assumptions & Constraints

• Assumptions: Things believed true but unverified (e.g., "Users have internet connectivity")
• Constraints: Hard limits on the solution space (budget, timeline, platform, regulatory)

10. Design Considerations (optional)

• UI/UX direction, rough wireframes, or key flows if they clarify intent
• Existing components to reuse
• Don't abdicate UX thinking, but don't over-prescribe either

11. Technical Considerations (optional)

• Known constraints, integration points, performance requirements
• Include only when genuinely relevant

12. Open Questions

Remaining unknowns or areas needing input. Better to surface these than hide them.

Anti-Patterns

• The "complete" but empty PRD — Every section filled with generic content that helps no one. A focused half-draft beats a template of platitudes.
• The solution masquerading as requirements — Describing database schemas instead of user outcomes.
• The novel — 20+ pages no one reads. Break large efforts into a parent PRD and child specs.
• Missing the other side — Only presenting upside without tradeoffs or risks.
• Delegating all thinking — "Design: TBD" with no direction whatsoever.
• No competitive context — Building without understanding what alternatives users have today.

Review Checklist

Before saving, verify:

• Could someone outside the planning meetings understand WHY we're building this?
• Are success metrics specific and measurable?
• Is scope clearly bounded (in and out)?
• Are user stories small, specific, and verifiable?
• Are functional requirements numbered and unambiguous?
• Does it address risks and tradeoffs honestly?
• Is it backed by evidence where possible?
• Would you be excited to build this after reading it?
• Is every section earning its place?

When NOT to Write a PRD

• Write a strategy doc / 1-pager instead when still exploring the opportunity space and validating hypotheses.
• Write a design/technical doc instead when the problem is understood but you need to work through the approach.