← Skills Library
Intelligence

PRD Writer

Idea to structured PRD in one command. Scope, requirements, measurable success criteria, and the open questions that will block engineering if you don't answer them first.

How It Works

PRD Writer · Workflow
Idea in, structured PRD out — markdown plus readable HTML.
TriggerIdea or research brief · A sentence, conversation, or /research-team output
1
Gather context
Target repo, CLAUDE.md, architecture, stakeholders, constraints
2
Diagrams first
Phase 0 — render-verified architecture + flow pack via /prd-diagrams
3
Write the markdown PRD
Problem · metrics · testable AC · tech spec · agent-team plan
4
Build HTML companion
Mission Control theme + interactive operator checklist + diagrams
5
Sanitization gateGATE
Leak guard on outward-facing files — PII / infra / codenames
failFindings → redact before any publish
PRD + handoff · Markdown + HTML to docs/prd/ · /ship command printed
  1. Idea or research brief: A sentence, conversation, or /research-team output
  2. Gather context: Target repo, CLAUDE.md, architecture, stakeholders, constraints
  3. Diagrams first: Phase 0 — render-verified architecture + flow pack via /prd-diagrams
  4. Write the markdown PRD: Problem · metrics · testable AC · tech spec · agent-team plan
  5. Build HTML companion: Mission Control theme + interactive operator checklist + diagrams
  6. Sanitization gate: Leak guard on outward-facing files — PII / infra / codenames
  7. PRD + handoff: Markdown + HTML to docs/prd/ · /ship command printed
ↆ download card

Invocation Triggers

/prd-writerwrite PRDproduct requirementsspec this featuredocument this idea

Use Cases

  • Turn a Slack idea into a structured PRD before the first line of code
  • Write explicit non-goals to kill scope creep before it starts
  • Surface open questions that will block engineering if left unresolved

The Problem

Product ideas die in Slack, or they survive as half-formed intentions that engineers have to interpret. The PRDs that do get written are either novels nobody reads or one-liners that leave too much to imagination. Engineers start without clarity and finish building the wrong thing.

What It Does

  1. 1
    Extract the core idea

    Who is the user, what is the problem, what is the proposed solution, why now — not "someday", but why this sprint. If the answer to "why now" is unclear, that is the first open question.

  2. 2
    Define scope

    Explicit in/out of scope and non-goals. Non-goals are the most important section: they prevent scope creep and resolve half the arguments before they start.

  3. 3
    Write functional requirements

    User-facing behaviors, not technical specs. "User can reset their password via email" not "implement SMTP with JWT token validation." The requirement describes what the system does for the user — the implementation is the engineer's domain.

  4. 4
    Write success criteria

    Measurable, specific, testable. Not "users should find it easy" — "time to complete the reset flow is under 60 seconds for 90% of users." If you cannot tell whether the criteria are met, they are too vague.

  5. 5
    Surface open questions

    Unresolved decisions that will block implementation. Who decides, and by when. If the PRD has no open questions, it is either perfect or incomplete. Almost always incomplete.

What You Get / What It Doesn't Do

What you get
  • Structured PRD with title, date, and author
  • Scope section with explicit non-goals
  • Functional requirements written as user-facing behaviors
  • Measurable success criteria that are specific and testable
  • Open questions block with owner and deadline for each
What it doesn't do
  • Make product decisions for you
  • Write technical specs or implementation plans
  • Validate with real users
  • Replace a discovery sprint for genuinely novel problems

Tips

The open questions section is more valuable than the requirements

It is where the real work is. A PRD with five solid open questions and ten requirements beats a PRD with zero open questions and thirty requirements — the latter is hiding unresolved decisions in prose.

A requirement without a named user is a wish list

"Users want X" is not a requirement. Name someone specific — a persona, a segment, a job title. That specificity forces real prioritization.

Read success criteria out loud

After writing, read each one out loud. If you cannot tell whether it is met after the feature ships, it is too vague. Rewrite until it is unambiguous.

Get the Skill

Pro SkillPRO

Unlock the full PRD Writer SKILL.md — drop it into ~/.claude/skills/ and trigger it by name.

What you unlock
  • Structured PRD with title, date, and author
  • Scope section with explicit non-goals
  • Functional requirements written as user-facing behaviors
  • Measurable success criteria that are specific and testable
...

Commonly Used With

Skills Library

29 more production skills ready to install.

Browse All Skills