Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate comprehensive Product Requirements Documents (PRDs) for product managers. Use this skill when users ask to "create a PRD", "write product requirements", "document a feature", or need help structuring product specifications.
.claude/skills/nicepkg-prd-generator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 113% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 82% | 0% |
| case-15 | ✓→✓ | = Same ✓ | 163% | 0% |
| case-21 | ✓→✓ | = Same ✓ | 129% | 0% |
| case-03 | ✓→✓ | = Same ✓ | 148% | 0% |
Generate comprehensive, well-structured Product Requirements Documents (PRDs) that follow industry best practices. This skill helps product managers create clear, actionable requirements documents that align stakeholders and guide development teams.
When a user requests to create a PRD (e.g., "create a PRD for a user authentication feature"), follow this workflow:
Before generating the PRD, collect essential information through a discovery conversation:
Required Information:
Discovery Questions to Ask:
1. What problem are you trying to solve?
2. Who is the primary user/audience for this feature?
3. What are the key business objectives?
4. Are there any technical constraints we should be aware of?
5. What does success look like? How will you measure it?
6. What's the timeline for this feature?
7. What's explicitly out of scope?Note: If the user provides a detailed brief or requirements upfront, you can skip some questions. Always ask for clarification on missing critical information.
Use the standard PRD template from references/prd_template.md to create a well-structured document. The PRD should include:
For each major requirement, generate user stories using the standard format:
As a [user type],
I want to [action],
So that [benefit/value].
Acceptance Criteria:
- [Specific, testable criterion 1]
- [Specific, testable criterion 2]
- [Specific, testable criterion 3]Reference references/user_story_examples.md for common patterns and best practices.
Use appropriate metrics frameworks based on the product type:
Consult references/metrics_frameworks.md for detailed guidance on each framework.
Optionally run the validation script to ensure PRD completeness:
bashscripts/validate_prd.sh <prd_file.md>
This checks for:
User Request: "Create a PRD for adding dark mode to our mobile app"
Execution:
User Request: "Write requirements for improving our search functionality"
Execution:
User Request: "I need a PRD for a new analytics dashboard product"
Execution:
User Request: "Create a lightweight PRD for a small bug fix feature"
Execution:
Good Requirements Are:
Avoid:
DO:
DON'T:
In-Scope Section:
Out-of-Scope Section:
Choose Metrics That:
Typical Metric Categories:
The skill supports different PRD formats:
Standard PRD - Full comprehensive document Lean PRD - Streamlined for agile teams One-Pager - Executive summary format Technical PRD - Engineering-focused requirements Design PRD - UX/UI-focused requirements
Specify the format when requesting: "Create a lean PRD for..." or "Generate a technical PRD for..."
Design Requirements Section Should Include:
Should Address:
PRD Should Help:
Distribution Checklist:
When creating a PRD based on customer feedback:
When creating a PRD for a strategic company initiative:
When creating a PRD for technical improvements:
When creating a PRD for compliance requirements:
Before finalizing the PRD, verify:
bash# Basic validation scripts/validate_prd.sh my_prd.md # Verbose output with suggestions scripts/validate_prd.sh my_prd.md --verbose # Check specific sections only scripts/validate_prd.sh my_prd.md --sections "user-stories,metrics"
This skill includes bundled resources:
bash# User: "Create a PRD for adding biometric authentication to our iOS app" # Assistant will: # 1. Ask discovery questions about security requirements, user personas, existing auth # 2. Generate PRD covering: # - Problem: Password friction, security concerns # - Solution: Face ID / Touch ID integration # - User stories: Enable biometric, fallback to password, settings management # - Metrics: Adoption rate, login success rate, support tickets # - Technical: iOS Keychain, LocalAuthentication framework # - Risks: Device compatibility, user privacy concerns # 3. Output formatted markdown PRD
bash# User: "Write requirements for improving our checkout flow conversion" # Assistant will: # 1. Gather data on current conversion rates and drop-off points # 2. Generate PRD including: # - Current state analysis with metrics # - Proposed improvements (guest checkout, saved payment, progress indicator) # - A/B test plan # - Success metrics: Conversion rate increase, time to checkout # - User stories for each improvement # 3. Include phased rollout approach
bash# User: "I need a PRD for an admin dashboard for enterprise customers" # Assistant will: # 1. Identify B2B-specific requirements (multi-tenancy, permissions, reporting) # 2. Generate comprehensive PRD with: # - Enterprise user personas (admin, manager, analyst) # - Role-based access control requirements # - Reporting and analytics needs # - Integration requirements (SSO, SCIM) # - Success metrics: Customer adoption, admin efficiency # 3. Include enterprise-specific considerations (compliance, SLAs)
Issue: PRD is too long/detailed
Solution: Create a "Lean PRD" focusing on problem, solution, acceptance criteria, and metrics. Reserve full PRD for major initiatives.
Issue: Requirements are too vague
Solution: Add specific examples, use concrete numbers, include visual references. Replace "fast" with "loads in under 2 seconds."
Issue: Stakeholders not aligned
Solution: Share PRD early as draft, incorporate feedback, present in person, get explicit sign-off before development starts.
Issue: Scope keeps expanding
Solution: Use "Out of Scope" section aggressively, create separate PRDs for future phases, tie scope to timeline constraints.
Issue: Engineers say it's not feasible
Solution: Involve engineering earlier in process, be flexible on solution approach, focus on problem not implementation.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 24,719 | 19,591 | -21% | 1 | 1 | 0% | 3,478 | 6,324 | +82% | 0 | 0 | — |
case-15 | pass→pass | 16,148 | 17,711 | +10% | 1 | 1 | 0% | 2,314 | 6,077 | +163% | 0 | 0 | — |
case-21 | pass→pass | 13,940 | 4,250 | -70% | 1 | 1 | 0% | 1,867 | 4,267 | +129% | 0 | 0 | — |
case-03 | pass→pass | 12,464 | 9,068 | -27% | 1 | 1 | 0% | 2,027 | 5,032 | +148% | 0 | 0 | — |
case-04 | pass→pass | 33,065 | 34,048 | +3% | 1 | 1 | 0% | 5,891 | 9,140 | +55% | 0 | 0 | — |
case-01 | fail→fail | 16,060 | 29,311 | +83% | 1 | 1 | 0% | 2,460 | 5,659 | +130% | 0 | 0 | — |
case-02 | fail→fail | 21,233 | 24,798 | +17% | 1 | 1 | 0% | 3,503 | 7,371 | +110% | 0 | 0 | — |
case-06 | pass→pass | 5,935 | 2,326 | -61% | 1 | 1 | 0% | 994 | 3,936 | +296% | 0 | 0 | — |
case-07 | pass→pass | 6,631 | 2,974 | -55% | 1 | 1 | 0% | 1,101 | 4,040 | +267% | 0 | 0 | — |
case-08 | pass→pass | 13,294 | 13,564 | +2% | 1 | 1 | 0% | 2,196 | 5,739 | +161% | 0 | 0 | — |
case-09 | pass→pass | 15,130 | 13,619 | -10% | 1 | 1 | 0% | 2,441 | 5,690 | +133% | 0 | 0 | — |
case-16 | pass→pass | 17,927 | 20,886 | +17% | 1 | 1 | 0% | 2,464 | 6,602 | +168% | 0 | 0 | — |
case-10 | pass→pass | 5,620 | 5,076 | -10% | 1 | 1 | 0% | 947 | 4,391 | +364% | 0 | 0 | — |
case-11 | pass→pass | 10,679 | 11,772 | +10% | 1 | 1 | 0% | 1,761 | 5,455 | +210% | 0 | 0 | — |
case-12 | pass→pass | 14,859 | 14,364 | -3% | 1 | 1 | 0% | 2,261 | 5,680 | +151% | 0 | 0 | — |
case-13 | pass→pass | 9,054 | 5,367 | -41% | 1 | 1 | 0% | 1,329 | 4,340 | +227% | 0 | 0 | — |
case-14 | fail→pass | 17,948 | 11,934 | -34% | 1 | 1 | 0% | 2,492 | 5,315 | +113% | 0 | 0 | — |
case-17 | pass→pass | 13,413 | 12,897 | -4% | 1 | 1 | 0% | 2,035 | 5,511 | +171% | 0 | 0 | — |
case-18 | pass→pass | 12,818 | 12,762 | -0% | 1 | 1 | 0% | 2,037 | 5,546 | +172% | 0 | 0 | — |
case-19 | pass→pass | 20,397 | 18,747 | -8% | 1 | 1 | 0% | 3,197 | 6,499 | +103% | 0 | 0 | — |
case-20 | pass→pass | 14,598 | 15,848 | +9% | 1 | 1 | 0% | 2,193 | 5,914 | +170% | 0 | 0 | — |
case-22 | pass→pass | 14,214 | 12,045 | -15% | 1 | 1 | 0% | 1,904 | 5,300 | +178% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.