PRD Writing
Turns rough notes into a structured PRD. Asks clarifying questions first, then writes. The blank-page problem is what this solves.
Staring at an empty Confluence template for 40 minutes. You know what the feature should do but can not make the first sentence come out right.
You braindump in plain language. PM Pilot asks 2 clarifying questions, then writes a structured PRD with problem, users, metrics, and open questions.
# PRD: Notification Preferences Redesign
# Problem
Users receive too many irrelevant notifications and have no granular control.
Support tickets citing notification fatigue up 34% this quarter.
# Target users
Power users managing 5+ projects (top 20% by DAU)
# Success metrics
• Notification-related support tickets down 30% in 60 days
• Settings completion rate above 70%
# Out of scope
• Email digest redesign (separate initiative)
• Mobile push notification channel
# Open questions [need answer before build]
• Do we support per-project or only global preferences?
What it does
You paste your rough thinking: a Slack thread, a Jira ticket, your half-formed notes. The skill asks 2-3 clarifying questions to force scope decisions, then writes a PRD covering problem, users, success metrics, out-of-scope, and open questions.
The output isn't a fill-in-the-template document. It's a PRD that reflects the actual decisions you made during the clarifying questions.
Why it asks questions first
The hardest part of a PRD is deciding what's in scope. Most people skip this and write a PRD that sounds complete but has hidden assumptions. Ask a stakeholder what "all users" means and you'll get five different answers.
/prd makes you answer the scope question before you touch formatting. The questions it asks are the ones your stakeholders will ask in review. Better to surface them now.
When to use it
Kicking off a new feature. You have a vague idea and a Jira ticket. Run /prd to go from "we should do X" to a document that engineering can react to.
After a discovery sprint. You've talked to users, you know the problem, you don't have the structure yet. Paste your research notes and let the skill find the shape.
Before an executive review. You have two days and need a PRD that won't embarrass you. Start here, then refine.
How it works
- You run
/prd for [feature] - The skill asks 2-3 clarifying questions about scope, target users, and success metric
- You answer, even briefly
- The skill writes the PRD
Try it
/prd for a notification preferences redesignExpected clarifying questions:
Before I write this, three things I need from you:
1. Who is the primary user: the person receiving notifications, or an admin
configuring them for a team?
2. What does "redesign" mean here: new UI only, or also a change to what
types of notifications are available?
3. What does success look like in 90 days: adoption rate, support ticket
volume, something else?Answer those and the PRD follows immediately.
Output structure
- Problem statement: what's broken and why it matters
- Target users: who specifically, not "all users"
- Success metrics: what you'll measure at 30/60/90 days
- Out of scope: explicit list of what you're not doing
- Open questions: the decisions still pending
What you need
Works with any setup, including zero install (copy-paste into claude.ai). No integrations required. The quality depends on how specific you are in the clarifying questions.
Pitfalls
Answering the clarifying questions vaguely. "Target users: all users" means the PRD will be vague too. If you say "all users," the skill will write for all users and the PRD will be useless in review.
Treating the first output as final. The PRD is a starting point. Share it with one engineer and one stakeholder before sending it to the room. Two rounds of feedback make it a real document.
Ready to install?
PM Pilot lives on GitHub. Every skill is a plain markdown file you can read, edit, and install in under 5 minutes.