prd（珠三角）

PRD

Create a structured Product Requirements Document with user stories, acceptance criteria, functional requirements, and scope boundaries.

When to Use

• A direction has been chosen and needs to be formalized into requirements
• Before architecture or task breakdown — requirements must exist first
• When the team needs a shared source of truth for what "done" means
• When acceptance criteria need to be explicit and verifiable

When NOT to use: Requirements are already documented in a prd.md that is still accurate, or the change is a single-line fix with obvious scope.

Input

• brainstorming.md from the artifact folder (required — contains chosen direction)
• goal-definition.md from the artifact folder (required — contains success criteria)
• context-map.md from the artifact folder (recommended — for technical grounding)

Output

• prd.md saved to the artifact folder (see references/formats.md for template)

The Process

Step 1: Load Upstream Artifacts

Read brainstorming.md, goal-definition.md, and context-map.md from the artifact folder. If brainstorming.md or goal-definition.md is missing, tell the user which skill to run first.

Also check for project documentation listed in references/context-sources.md (docs/TechStack.md, docs/ProjectStructure.md). If available, use them to inform technical considerations.

Extract:

• The chosen direction and rationale (from brainstorming)
• Success criteria and constraints (from goal definition)
• Tech stack and existing patterns (from context map)

Step 2: Ask Clarifying Questions

Ask 3-5 essential clarifying questions where the chosen direction leaves gaps. Focus on:

• Core Functionality: What are the key actions a user takes?
• Scope Boundaries: What should this specifically NOT do?
• Edge Cases: What happens when things go wrong?
• Success Criteria: How do we measure "done" beyond the goal definition?

Format questions with lettered options for quick answers:

1. How should the system handle duplicate entries?
   A. Allow duplicates silently
   B. Reject with a validation error
   C. Warn the user and let them decide
   D. Auto-merge duplicates

2. What is the scope for the first version?
   A. Minimal viable — core happy path only
   B. Full-featured with error handling
   C. Core path + most common edge cases
   D. Other: [please specify]

Users can answer with "1C, 2A" for quick iteration.

Only ask questions where the answer genuinely changes the PRD. Do not ask questions you can answer from the upstream artifacts.

Step 3: Write User Stories

For each piece of functionality, write a user story:

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

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

Rules for user stories:

• Each story must be small enough to implement in one focused session
• Acceptance criteria must be verifiable, never vague
• Bad: "Works correctly", "Good UX", "Handles edge cases"
• Good: "Button shows confirmation dialog before deleting", "Response returns within 500ms"
• Number stories sequentially: US-001, US-002, etc.

Step 4: Write Functional Requirements

Number each requirement for easy reference:

- FR-1: The system must allow users to create an account with email and password
- FR-2: When a user submits an invalid email, the system must display an inline error message
- FR-3: Passwords must be at least 8 characters with one number and one special character

Be explicit and unambiguous. The reader may be a junior developer or AI agent.

Step 5: Define Non-Goals

Carry over the "Not Doing" list from goal definition and expand it based on the clarifying questions:

## Non-Goals (Out of Scope)
- No email verification in v1 — will be added later
- No social login — only email/password for now
- No admin dashboard — management via database directly

This section prevents scope creep. Every PRD must have non-goals.

Step 6: Add Technical and Design Considerations

Based on the context map, note:

• Existing components that can be reused
• Integration points with existing systems
• Performance requirements
• Known constraints or dependencies

Step 7: Define Success Metrics

Translate the success criteria from goal definition into measurable metrics:

## Success Metrics
- Registration completes in under 3 clicks
- Error rate for form submission < 5%
- Page load time < 2 seconds

Step 8: Write and Save PRD

Write prd.md to the artifact folder using the template from references/formats.md. Include all sections from Steps 3-7, plus an Open Questions section for anything unresolved.

Present the PRD to the user for review before saving. Apply any requested changes.

Announce the saved path:

"PRD saved to [path]/prd.md."

Writing Quality

The PRD reader may be a junior developer or AI agent:

• Be explicit and unambiguous
• Avoid jargon or explain it
• Provide concrete examples where helpful
• Number requirements for easy reference
• Make acceptance criteria verifiable (not "works correctly")

Common Rationalizations

Red Flags

• User stories without acceptance criteria
• Acceptance criteria that say "works correctly" or "is intuitive"
• No non-goals section
• No functional requirements — only user stories (or vice versa)
• PRD that doesn't reference the chosen direction from brainstorming
• Proceeding to architecture without user reviewing the PRD

Verification

Before handing off, confirm:

• Upstream artifacts loaded (brainstorming, goal definition)
• 3-5 clarifying questions asked where needed
• Every user story has verifiable acceptance criteria
• Functional requirements are numbered and unambiguous
• Non-goals section has at least 2 items
• Success metrics are measurable
• User has reviewed and approved the PRD
• prd.md saved to artifact folder

Next

"PRD complete. Next recommended skill: architecture — design the technical approach for these requirements."