PRD Writer — Claude Skill

PRDs your engineers
can actually
build from.

Full PRD. GWT criteria. Edge cases. No rework.

Feature idea in, engineering-ready PRD out — user stories, acceptance criteria, NFRs, and success metrics included.

PRD Writer is a Claude AI skill that converts a feature idea into a complete, engineering-ready PRD — user stories, GWT acceptance criteria, edge cases, NFRs, and success metrics.

⚡ Replaces 4–6 hours of requirements writing — runs in under 5 minutes.
Get this skill — $15
Output preview — PRD Writer
Problem Statement
Outcome-first framing. Problem only — no solution leak.
Current state: Admins manually export CSVs from three separate dashboards, merge them in Sheets, and email the result to stakeholders — typically taking 45–90 minutes each Monday.

Pain point: The process breaks when any one export format changes. 3 of the last 8 Mondays resulted in an error that delayed the report by half a day.

Why now: Q3 headcount growth means the manual process will hit its limit within 60 days.
Under 100 words. Scannable in 10 seconds. No solutions in this section.
User Story + GWT Criteria
Behaviour-first. QA-testable without asking the PM.
US-02: Scheduled report delivery
Priority: Must Have

As an admin, I want weekly reports delivered automatically to stakeholder inboxes, so that I don't have to manually compile and send them.

Scenario 1 — Successful delivery
Given a report schedule is configured and active
When Monday 08:00 UTC arrives
Then stakeholders receive the report within 5 minutes

Scenario 2 — Delivery failure
Given an email address in the list is invalid
When the report is dispatched
Then the admin receives a failure notification within 10 minutes
Every Must Have story has a failure scenario. Acceptance criteria need no PM clarification.
Edge Cases
Five categories covered. Assumptions flagged explicitly.
Empty state: No data in selected date range → show "No data available for this period" with a date-reset CTA
Auth failure: User session expires mid-export → redirect to login, restore export context on return
Network error: Export job times out after 30s → surface retry button, do not silently fail
Concurrent action: Two admins trigger export simultaneously → queue second request, do not duplicate output
Data limit: Export exceeds 50,000 rows → prompt user to apply date filters, block oversized export [ASSUMED — please confirm limit]
Minimum 5 edge cases per PRD. Assumptions flagged as [ASSUMED] for PM sign-off.
Non-Functional Requirements
Quantified. Testable. No "fast" or "secure" allowed.
Performance: Export job completes in <8s for datasets up to 10,000 rows at p95
Security: Exported files encrypted in transit (TLS 1.3); download links expire after 24 hours
Reliability: Scheduled reports must achieve 99.5% on-time delivery; failures trigger admin alert within 10 min
Accessibility: Export controls meet WCAG 2.1 AA
Compliance: GDPR — exported data must not include fields flagged as PII unless admin has data-export role [NEEDS TECH REVIEW]
Every NFR is measurable. Items needing engineering input flagged as [NEEDS TECH REVIEW].
Full output
What you receive
✓ Problem statement — outcome-first, under 100 words
✓ Goals & success metrics table with baseline + target
✓ User personas (primary + secondary)
✓ 3–8 user stories — As a / I want / So that format
✓ Given-When-Then acceptance criteria per story
✓ Edge cases table — 5+ rows across all failure categories
✓ Non-functional requirements — quantified, testable
✓ Out-of-scope section + open questions log
✓ Ready to paste into Notion, Linear, Confluence, or Jira
1 / 5
How it works

How does PRD Writer work?

  1. Tell Claude what you need — the skill asks 2-3 narrowing questions specific to the task.
  2. The skill runs live research and structures the output before writing a single word.
  3. An anti-slop quality gate checks the draft, then delivers a ready-to-use result for $15.
What you get

Why engineering
actually builds it.

📐
GWT acceptance criteria — QA-testable without asking the PM.
Zero-ambiguity spec
🔍
Edge cases across 5 failure categories — no surprises in sprint.
Bug prevention
📊
Success metrics with baseline and target — every feature measured.
Outcome-first
📋
Paste-ready for Notion, Linear, Confluence, or Jira. No reformatting.
4-tool compatible
Who it's for

Built for anyone
shipping product.

✓
PMs at startups writing PRDs solo, without a PM team.
✓
Founders who need engineering-ready specs from rough ideas.
✓
Senior PMs reducing rework cycles with faster first drafts.
✓
Product teams using Linear, Jira, Confluence, or Notion.
Not what this is for
✕Technical design or architecture docs — this covers requirements, not implementation.
✕Full epics with 15+ user stories — features that large should be split first.
✕Non-software projects — PRD structure is built for product and engineering teams.

The complete guide to writing prds with ai covers what makes a spec engineers can act on versus one that needs a follow-up meeting.

The complete guide to writing prds with ai covers what makes a spec engineers can build from versus one that needs a follow-up meeting.

"First PRD I've written where engineering didn't come back with a single clarifying question. Zero back-and-forth before sprint start."
KT
Kenji T.
Senior product manager · Tokyo
Get it now
The PRD
engineering ships.
$15.

Instant download. Works with your free Claude account. 7-day refund if it's not what you expected.

🛡️ 7-day money back guarantee · no questions asked
Quick answers

Do I need a paid Claude account for PRD Writer?

No. The free Claude plan is enough to get started and see the skill working. A Pro plan unlocks longer outputs.

How do I install the PRD Writer skill?

Download the .skill file, open Claude, go to Customize → Skills → Upload a skill, and drop the file in. It appears in your slash-command list immediately.

Can I edit or customise the output after I receive it?

Yes. The output is plain text you own and can edit freely. Most users tweak specific details or the call-to-action to fit their exact use case before sending or publishing.

How specific do my inputs need to be to get a good result?

The more specific your input, the more precise the output. The skill asks targeted questions to gather context — answering these in detail produces noticeably better results than generic one-line inputs.

Free skill
Try NovaKit before
you spend a dollar.

Get the LinkedIn Post Engine free — the same skill that runs live trend research before every post. Drop your email and it lands in your inbox in seconds.

💼
LinkedIn Post Engine
Social · normally $9 · free today
✓Live trend research before every post
✓Hook variants calibrated to what's converting this week
✓Works on a free Claude account

No spam. No account. Unsubscribe any time.