An interactive Claude Code skill that guides you through writing production-grade PRDs. Section by section, with built-in validation, user story, and Gherkin acceptance criteria.
Built on the Common PRD Standards — 12 core sections, 4 optional sections, full rules per section.
- Runs an upfront intake to gather product type, platform, team structure, compliance exposure, and discovery maturity before any section begins — so every subsequent question is targeted and context-aware
- Identifies applicable regulations when compliance signals are present (payments, eKYC, PII, regulated data) — maps them to GDPR, PCI-DSS, HIPAA, OJK, and others, confirms the list with you, and propagates them into Background, Scope, and Risks automatically
- Interviews you section by section with focused, contextual questions
- Validates every answer against the section rules before moving on
- Enforces user story format for all requirements
- Generates Gherkin acceptance criteria automatically
- Never fabricates content — if you don't have an answer, it writes
[TBD]instead of inventing metrics, names, dates, or evidence - Scores the completed PRD out of 100 with a detailed validation report
- Outputs a production-ready
docs/prd.mdfile
Terminal (developers):
git clone https://github.com/anatasof/NatPRD.git ~/.claude/skills/NatPRDRestart Claude Code — the skill loads automatically.
No terminal (everyone else):
- Go to github.com/anatasof/NatPRD
- Click the green Code button → Download ZIP
- Unzip it and rename the folder to
NatPRD - Move it into your Claude skills folder:
- Mac: Open Finder → press
⌘ Shift G→ paste~/.claude/skills→ drop the folder in - Windows: Navigate to
C:\Users\[your name]\.claude\skills\
- Mac: Open Finder → press
- Restart Claude Code
The skill loads automatically. Just say what you need and Claude handles the rest.
No native skill install — load it manually instead:
Option A — Project instructions (recommended, available on all platforms):
- Create a new Project in the Claude app
- Open the project → click Edit project instructions
- Copy the full contents of
SKILL.mdfrom this repo and paste it in - Save — every conversation in that project follows the PRD workflow
Option B — Customize Claude (applies to all your chats):
- Click your profile → Customize Claude
- Paste the contents of
SKILL.mdinto the instructions field - Save
Note: The Claude app has no file system access. Claude generates the PRD content in chat — copy it and save it yourself as docs/prd.md.
Start Claude Code in any project directory, then use natural language:
# Create a new PRD
"I want a PRD for adding two-factor authentication"
"Help me write a PRD for the merchant onboarding flow"
"Create a PRD for our loyalty points system"
# Review an existing PRD
"Review my PRD"
"Validate the PRD at docs/prd.md"
# Update a specific section
"Update the hypothesis in the PRD"
"Add a new user story to the requirements section"
# Generate a stakeholder summary
"Generate a summary of the PRD for stakeholders"
- Initiative Name
- Document Status ← status, version, author, owner, reviewers, approvers, approval date
- Background
- Objective
- Scope & Boundaries
- Hypothesis
- Success Metrics
- Requirements ← user stories + Gherkin AC + NFRs
- Solution
- Metric Monitoring
- Event & Data Tracking
- FAQ
- Risks & Mitigations — regulatory risk rows (one per confirmed regulation) + operational risks; legal sign-off checkpoint included
- Dependencies — for cross-team or third-party work
- Launch Plan — for coordinated rollouts
- Stakeholder Map — for cross-functional initiatives
After generation, the PRD is scored out of 100:
| Band | Score | Meaning |
|---|---|---|
| Excellent | 90–100 | Ready for In Review |
| Good | 75–89 | Fix warnings before circulating |
| Needs Work | 60–74 | Several sections incomplete |
| Not Ready | <60 | Major gaps — do not circulate |
NatPRD/
├── SKILL.md ← Skill entry point (Claude Code reads this)
├── README.md ← This file
├── templates/
│ ├── prd-template.md ← Blank PRD template
│ └── prd-summary-template.md ← One-page stakeholder summary
├── prompts/
│ ├── interview-questions.md ← Questions asked per section during generation
│ ├── section-rules.md ← Full rules per section (used in validation)
│ └── validation-rules.md ← Scoring rubric and report format
| Mode | Trigger | What Happens |
|---|---|---|
| Generate | "I want a PRD for..." | Upfront intake → guided section interview → optional sections → full PRD output |
| Review | "Review my PRD" | Reads existing file → validates → scores → report |
| Update | "Update [section]" | Targets one section → re-interviews → rewrites → re-validates |
When a compliance signal is confirmed at intake (payments, eKYC, KYB, PII, biometrics, healthcare, or cross-border data), the skill runs a dedicated regulation identification step before the main interview. It presents a signal-to-regulation mapping (GDPR, PCI-DSS, HIPAA, OJK, PDPA, and more), confirms the list with you, and carries the confirmed regulations forward automatically:
- Background — adds a Regulatory Context table with jurisdiction, key obligations, and review status
- Scope — flags regulatory-driven in/out-of-scope items separately so reviewers can trace them
- Risks & Mitigations — adds one risk row per confirmed regulation with mitigation, contingency, named owner, and a legal/compliance sign-off checkbox
If you don't know which regulations apply, the skill writes [TBD — legal/compliance to confirm] and flags it as an open item rather than inventing requirements.
- PRD file:
docs/prd.md - Validation report: Printed inline after generation
- Summary (optional):
docs/prd-summary.md
Q: Does this work outside Claude Code?
A: Yes, with a manual step. Paste the contents of SKILL.md into a Project's instructions on Claude.ai (web or mobile), or into the Customize feature. The PRD workflow works the same way — the only difference is Claude cannot save files for you, so you copy the output yourself.
Q: Why does it ask so many questions upfront before the PRD sections start? A: The intake phase gathers context about your product type, platform, team structure, compliance exposure, and discovery maturity. Without it, Claude would have to guess — and guessing leads to invented metrics, fabricated design links, and hallucinated evidence. The intake makes every subsequent question more specific and the output more accurate.
Q: Can I customise the template?
A: Yes. Edit templates/prd-template.md and prompts/section-rules.md to match your team's standards.
BSD 3-Clause — free to use and redistribute, with attribution. Do not use the author's name to endorse derived works.