Prd

Create a Product Requirements Document for: $ARGUMENTS

Context Awareness

Before starting, check what already exists:

• Read docs/ directory structure if it exists
• Read any existing PRD, README, or whitepaper — do not duplicate what's already written
• If a PRD already exists, ask the user if they want to revise it or start fresh

Process

1. Deep Interview

Use AskUserQuestion to interview the user. This is the most important step — the PRD quality depends entirely on how well you extract the user's thinking.

Start broad, then go deep:

1. The problem — What problem are you solving? Who has this problem? How are they solving it today? What's broken about the current approach?
2. The solution — What's your high-level approach? What makes it different from existing solutions? What's the core insight?
3. Scope and boundaries — What does v1 include? What is explicitly out of scope? What are the non-negotiable requirements vs nice-to-haves?
4. Users and stakeholders — Who are the primary users? Are there different user types with different needs? Who else cares about this (ops, security, compliance)?
5. Success criteria — How do you know this works? What does success look like in 1 month? 6 months?
6. Constraints — Technical constraints (language, platform, infra)? Business constraints (timeline, team size, budget)? Regulatory constraints?
7. Risks — What could go wrong? What are the biggest unknowns? What keeps you up at night about this?

Interview rules:

• Don't ask obvious questions — if the user said "Rust CLI tool", don't ask "what language?"
• Dig into contradictions and tensions — "you said X but also Y, how do those reconcile?"
• Ask about the hard parts the user might not have thought through
• It's OK to push back or challenge assumptions — the goal is a solid PRD, not agreement
• Keep going until you have enough to write a comprehensive document. Don't rush.

2. Write the PRD

Write to docs/PRD.md (or docs/prd/.md if this is a sub-feature PRD).

# [Project/Feature Name]

## Problem

[What problem does this solve? Who is affected? What's the current state and why is it broken?]

## Solution

[High-level approach. What is this thing? What's the core insight? How does it work at a conceptual level?]

## Design Principles

[Non-negotiable properties of the solution. The "even if everything else changes, these must hold" rules.]

## Scope

### In Scope (v1)
- [Concrete deliverable 1]
- [Concrete deliverable 2]

### Out of Scope
- [Explicit exclusion 1 — and why]
- [Explicit exclusion 2 — and why]

## User Types

### [User Type 1]
- Who they are
- What they need
- How they interact with the system

## Key Flows

[The 2-4 most important user/system flows. Use prose, diagrams (Mermaid), or step-by-step — whatever is clearest.]

## Success Criteria

- [ ] [Measurable criterion 1]
- [ ] [Measurable criterion 2]

## Constraints

- **Technical:** [language, platform, infra, dependencies]
- **Business:** [timeline, team, budget]
- **Regulatory:** [compliance, security, legal]

## Risks and Open Questions

| Risk/Question | Impact | Mitigation/Answer |
|---------------|--------|-------------------|
| [Risk 1] | [What happens if it materializes] | [How to handle it] |

## Future Work

[What comes after v1? What was deferred and why?]

Adapt the structure to fit the project. Not every section is needed for every PRD. A small feature PRD might skip User Types and Constraints. A protocol PRD might need a Terminology section. Use judgment.

3. Present for Review

Show the user the PRD and ask for feedback. Iterate until they're satisfied.

4. Suggest Next Steps

Based on the PRD complexity, suggest which specs to create next:

• "This needs an architecture doc — run /architecture"
• "Define testing and dev workflow — run /tdd"
• "There are security-sensitive parts — run /security"
• "Ready to break into features — run /spec for each"

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: 0xrafasec
• Source: 0xrafasec/ai-workflow
• License: MIT

Install and usage instructions live in the source repository linked above.