SuperAPI AI Directory

prd

Coding & development

View repository

Updated 10-15 21:53

About this skill

Product Requirements Document (PRD)

Generate comprehensive, production-grade PRDs that bridge the gap between business vision and technical execution.

When to Use This Skill

Trigger when:

  • User wants to "write a PRD", "create product requirements", or "document a feature"
  • Transitioning from Intake to PRD solidification phase
  • User provides initial ideas, demos, or prototypes needing formalization
  • User asks to "plan a feature", "define scope", or "clarify requirements"

Produces:

  • prd/PRD.md - the main Product Requirements Document

Workflow

Decision Point: Discovery or Draft?

Ask clarifying questions FIRST when:

  • User provides only vague idea ("build something cool")
  • Critical details missing (who, what success looks like)
  • Multiple reasonable interpretations possible

Skip to drafting when:

  • User provides sufficient detail in initial request
  • Clear problem statement + target users + success criteria provided
  • User explicitly says "just write it" or provides complete context

Phase 1: Discovery (If Needed)

Gather context through targeted questions. Keep it brief - 3-5 questions max.

Typical Questions:

  • Problem: What pain point are we solving?
  • Why now: Why important at this time?
  • Success: How will we measure success?
  • Scope: What's in/out of scope?
  • Users: Who are the target users?

Phase 2: Analysis & Synthesis

Synthesize understanding:

  • Map key user flows
  • Identify core business processes
  • Define module boundaries
  • List key pages/scenes
  • Identify risks and dependencies

Phase 3: PRD Drafting

Generate document using the standard schema below, then:

  1. Write the completed PRD to prd/PRD.md (create prd/ directory if needed)
  2. Report the file path to the user
  3. Ask if any section needs adjustment

PRD Schema (Use These Exact Section Names)

1. Background & Goals

Must include:

  • Problem statement (what pain point?)
  • Why now (why important at this time?)
  • Success criteria (3-5 measurable KPIs)

2. User Personas

Must include:

  • Primary users identified
  • Their characteristics/jobs
  • Current workflow pain points

3. Core Business Processes

Must include:

  • Primary workflow steps
  • Decision points
  • Key user interactions

DO NOT include: Detailed state machines, API calls, database schemas

4. Module Boundaries

Must include:

  • What modules/components involved
  • How modules interact (high-level)
  • External dependencies

DO NOT include: Technical implementation details, code structure

5. Pages & Scenes Inventory

Must include:

  • List of key pages/scenes
  • Entry conditions
  • Exit conditions

6. Success Criteria

Must include:

  • Quantitative metrics (not vague "fast", "easy")
  • How each metric measured

Example:

  • BAD: "The system should be fast"
  • GOOD: "Search returns within 200ms for 10k records"

7. Scope / Non-Scope

Must include:

  • What's built in this phase
  • What's NOT built (explicitly)

8. Risks & Dependencies

Must include:

  • Technical risks
  • External dependencies
  • Key assumptions

Quality Standards

Include (Do)

  • Business context and motivation
  • User roles and workflows
  • Clear measurable success metrics
  • Explicit scope boundaries
  • Known risks and dependencies

Exclude (Don't)

These belong in SPEC, NOT PRD:

  • Field types or data structures
  • Database table designs
  • API response structures
  • Internal state machines
  • Detailed error codes
  • Technical architecture details
  • Code-level implementation

Example: What Goes Where

Content PRD SPEC
Problem statement ✅ ✅
User personas ✅ ✅
User flows ✅ ✅
Success metrics ✅ ✅
Module boundaries ✅ ✅
Page inventory ✅ ❌
API contracts ❌ ✅
Database schema ❌ ✅
State machines ❌ ✅
Technical architecture ❌ ✅

Expert Insights

"The most important section is the first part - what is the background and context? What is the problem, why does it matter, and why does it matter now?" - Maggie Crowley

"Whenever we're devising a new product, we start by writing a press release describing it in a way that speaks to the customer." - Bill Carr

"We tend to keep them pretty light. I like to have the minimal amount of context that ensures everyone's on the same page." - Eric Simons

"If you're not prototyping and building to see what you want to build, you're doing it wrong." - Aparna Chennapragada


Common Mistakes to Avoid

  1. Starting with solution - Always lead with problem and context
  2. No success criteria - Every PRD needs measurable KPIs
  3. Including technical details - Save API/schema for SPEC phase
  4. Vague scope - Explicitly state what's NOT included
  5. Missing "Why Now" - Justify timing, not just what and how
  6. Over-detailing - Keep PRD lightweight; save depth for SPEC

Completion Criteria

Ensure at completion:

  • Background clearly states problem and why now
  • User personas identified
  • Core business processes documented
  • Module boundaries defined (high-level only)
  • Pages/scenes inventory complete
  • Success criteria are measurable
  • Scope/non-scope explicitly listed
  • Risks and dependencies documented
  • NO technical implementation details

Output Location

Default to prd/PRD.md

Follow user's specified location if provided.

Always write to disk. Do not output the PRD only in the conversation.

Related Skills

  • Upstream — project-intake supplies docs/00-intake.md; run interview-me first when the business direction is still unresolved.
  • Downstream — spec-writer is always the next stage; when prototypes or demos exist, ux-analyst comes first. Hand off once the PRD completion criteria are met and no implementation details leaked in.
  • Boundary — use spec-writer for engineering contracts, state machines, and API or schema detail; this skill owns problem framing, personas, scope, and success metrics. Use ux-analyst for per-screen behavior semantics.
  • Companion — interview-me when discovery answers stay vague; if it is unavailable, list the open questions under risks and dependencies instead.
  • If a referenced skill is not installed in this session, name it in the handoff message and continue with the current artifact; do not stall.