PRD Generator Skill

Generate comprehensive Product Requirements Documents with templates, technical specs, success metrics, and launch readiness checklists.

Overview

This skill provides comprehensive capabilities for creating and managing Product Requirements Documents throughout the product development lifecycle. It transforms feature specifications into structured PRDs ready for cross-functional review.

Capabilities

PRD Generation

Generate PRD from feature specifications
Apply company templates and formatting
Create multiple PRD variants (detailed, executive summary)
Support incremental PRD development

Technical Specifications

Generate technical requirements sections
Define API contracts and data models
Specify integration requirements
Document technical constraints

Success Metrics

Define success criteria and KPIs
Create metric specifications
Set launch and long-term targets
Specify measurement methodology

Launch Planning

Build launch checklists
Define go/no-go criteria
Create rollout plans
Document rollback procedures

Version Management

Track PRD changes over time
Maintain revision history
Support collaborative editing
Generate change summaries

Prerequisites

Template Requirements

Supported templates:
- Standard PRD
- Technical PRD (API/Platform)
- Growth Feature PRD
- MVP/Experiment PRD
- Custom templates

Input Format

{
  "feature": {
    "name": "Feature name",
    "description": "Brief description",
    "problem_statement": "Problem being solved",
    "target_users": ["persona1", "persona2"]
  },
  "template": "standard",
  "sections": ["all"] // or specific sections
}

Usage Patterns

Standard PRD Template

# Product Requirements Document

## Document Info
| Field | Value |
|-------|-------|
| Feature Name | [Name] |
| Author | [PM Name] |
| Created | [Date] |
| Status | [Draft/Review/Approved] |
| Version | [1.0] |

---

## 1. Overview

### 1.1 Problem Statement
[What problem are we solving? Why now?]

### 1.2 Opportunity
[Market opportunity, business impact]

### 1.3 Target Users
[Primary and secondary user personas]

---

## 2. Goals & Success Metrics

### 2.1 Objectives
| Objective | Metric | Target | Timeline |
|-----------|--------|--------|----------|
| [Objective 1] | [Metric] | [Target] | [When] |

### 2.2 Non-Goals
- [What we explicitly won't do]

---

## 3. User Stories & Requirements

### 3.1 User Stories
[List of user stories with acceptance criteria]

### 3.2 Functional Requirements
| ID | Requirement | Priority | Notes |
|----|-------------|----------|-------|
| FR-001 | [Requirement] | P0 | [Notes] |

### 3.3 Non-Functional Requirements
- Performance: [Requirements]
- Security: [Requirements]
- Accessibility: [Requirements]

---

## 4. Solution Design

### 4.1 User Experience
[UX flow description, wireframe links]

### 4.2 Technical Approach
[High-level technical approach]

### 4.3 Dependencies
| Dependency | Team | Status |
|------------|------|--------|
| [Dependency] | [Team] | [Status] |

---

## 5. Launch Plan

### 5.1 Rollout Strategy
[Phased rollout plan]

### 5.2 Go/No-Go Criteria
- [ ] [Criterion 1]
- [ ] [Criterion 2]

### 5.3 Rollback Plan
[How to rollback if needed]

---

## 6. Open Questions
| Question | Owner | Due Date | Resolution |
|----------|-------|----------|------------|
| [Question] | [Owner] | [Date] | [Resolution] |

---

## Appendix
### A. Research & Data
### B. Competitive Analysis
### C. Technical Specifications
### D. Revision History

PRD Generation Example

## Input: Feature Specification

**Feature**: Smart Search with AI Suggestions
**Problem**: Users struggle to find relevant content, leading to frustration and abandonment.
**Target Users**: All registered users, especially new users.

---

## Generated PRD

# Product Requirements Document: Smart Search with AI Suggestions

## Document Info
| Field | Value |
|-------|-------|
| Feature Name | Smart Search with AI Suggestions |
| Author | [Generated] |
| Created | 2026-01-24 |
| Status | Draft |
| Version | 0.1 |

---

## 1. Overview

### 1.1 Problem Statement

Users currently struggle to find relevant content in our application:
- **40% of searches return zero results** (Analytics, Q4 2025)
- **Average 3.2 search attempts** before finding desired content
- **25% abandonment rate** when first search fails

This friction leads to user frustration, reduced engagement, and increased churn.

### 1.2 Opportunity

By implementing AI-powered search suggestions:
- Reduce zero-result searches by 60%
- Improve search success rate to first-try resolution
- Increase content discovery and engagement
- Differentiate from competitors with smarter search UX

**Business Impact**: Estimated 15% improvement in D30 retention based on correlation analysis.

### 1.3 Target Users

**Primary**: New users (0-30 days)
- Unfamiliar with content organization
- Need guidance on what's available
- Highest abandonment on search failure

**Secondary**: Power users
- Expect efficient search
- Value time savings
- Will adopt advanced features

---

## 2. Goals & Success Metrics

### 2.1 Objectives

| Objective | Metric | Target | Timeline |
|-----------|--------|--------|----------|
| Reduce search friction | Zero-result rate | < 15% | Launch + 30d |
| Improve findability | First-try success | > 70% | Launch + 30d |
| Increase engagement | Search-to-content clicks | +25% | Launch + 60d |
| Improve retention | D30 retention (searchers) | +5% | Launch + 90d |

### 2.2 Non-Goals

- We will NOT replace the existing basic search (this is an enhancement)
- We will NOT personalize suggestions in V1 (future iteration)
- We will NOT support voice search in this release

---

## 3. User Stories & Requirements

### 3.1 User Stories

**US-001: Search Suggestions**
As a user searching for content,
I want to see relevant suggestions as I type,
So that I can find what I'm looking for faster.

**Acceptance Criteria**:
- [ ] Suggestions appear after 2 characters typed
- [ ] Maximum 5 suggestions displayed
- [ ] Suggestions update within 200ms of typing
- [ ] Clicking suggestion executes search

**US-002: Typo Correction**
As a user who makes typos,
I want the search to understand my intent,
So that I still find relevant results.

**Acceptance Criteria**:
- [ ] System suggests corrections for misspellings
- [ ] "Did you mean..." shown when no exact matches
- [ ] User can click to search corrected term

### 3.2 Functional Requirements

| ID | Requirement | Priority | Notes |
|----|-------------|----------|-------|
| FR-001 | Display up to 5 search suggestions | P0 | Ranked by relevance |
| FR-002 | Support keyboard navigation of suggestions | P0 | Arrow keys, Enter |
| FR-003 | Show suggestion type (content, category) | P1 | Visual indicator |
| FR-004 | Handle typos with fuzzy matching | P1 | Levenshtein distance < 2 |
| FR-005 | Cache popular suggestions | P2 | Performance optimization |

### 3.3 Non-Functional Requirements

**Performance**:
- Suggestions load in < 200ms (p95)
- Search index updates within 5 minutes of content changes

**Security**:
- Suggestions respect user permissions
- No PII in suggestion logs

**Accessibility**:
- WCAG 2.1 AA compliant
- Screen reader support for suggestions

---

## 4. Solution Design

### 4.1 User Experience

[Wireframe link: /designs/smart-search-v1]

**Flow**:
1. User clicks search box
2. Recent searches shown (if any)
3. User types, suggestions appear after 2 chars
4. Suggestions update with each keystroke
5. User selects suggestion or presses Enter
6. Results displayed with matched terms highlighted

### 4.2 Technical Approach

**Search Infrastructure**:
- Elasticsearch with suggestion capability
- Prefix matching with boosted scoring
- Typo tolerance via fuzzy matching

**API**:
- New endpoint: `GET /api/search/suggest?q={query}`
- Response time SLA: 200ms p95

### 4.3 Dependencies

| Dependency | Team | Status |
|------------|------|--------|
| Elasticsearch upgrade | Platform | Scheduled Q1 |
| Search UI components | Design System | Available |
| Analytics tracking | Data | Needs spec |

---

## 5. Launch Plan

### 5.1 Rollout Strategy

1. **Week 1**: Internal dogfooding (employees)
2. **Week 2**: 5% of users (random sample)
3. **Week 3**: 25% of users (if metrics positive)
4. **Week 4**: 100% rollout

### 5.2 Go/No-Go Criteria

- [ ] p95 latency < 200ms in staging
- [ ] Zero P0 bugs in 5% rollout
- [ ] Suggestion relevance > 80% (sampled review)
- [ ] No increase in error rates

### 5.3 Rollback Plan

1. Disable feature flag `smart_search_suggestions`
2. Revert to basic search (no code change needed)
3. Notify users of temporary change

---

## 6. Open Questions

| Question | Owner | Due Date | Resolution |
|----------|-------|----------|------------|
| ML model vs rule-based for V1? | [Tech Lead] | 2026-01-28 | |
| How to handle multi-language? | [PM] | 2026-01-30 | |

---

## Appendix

### A. Research & Data
- User interview synthesis: [Link]
- Search analytics dashboard: [Link]
- Competitive analysis: [Link]

### B. Revision History
| Version | Date | Author | Changes |
|---------|------|--------|---------|
| 0.1 | 2026-01-24 | [Generated] | Initial draft |

Integration with Babysitter SDK

Task Definition Example

const prdGenerationTask = defineTask({
  name: 'generate-prd',
  description: 'Generate PRD from feature specification',

  inputs: {
    feature: { type: 'object', required: true },
    template: { type: 'string', default: 'standard' },
    sections: { type: 'array', default: ['all'] },
    includeTechnical: { type: 'boolean', default: true }
  },

  outputs: {
    prd: { type: 'string' },
    metadata: { type: 'object' },
    openQuestions: { type: 'array' }
  },

  async run(inputs, taskCtx) {
    return {
      kind: 'skill',
      title: `Generate PRD for ${inputs.feature.name}`,
      skill: {
        name: 'prd-generator',
        context: {
          operation: 'generate',
          feature: inputs.feature,
          template: inputs.template,
          sections: inputs.sections,
          includeTechnical: inputs.includeTechnical
        }
      },
      io: {
        inputJsonPath: `tasks/${taskCtx.effectId}/input.json`,
        outputJsonPath: `tasks/${taskCtx.effectId}/result.json`
      }
    };
  }
});

Output Formats

PRD Export Options

Formats:
  - Markdown (.md)
  - Google Docs (via API)
  - Confluence (via API)
  - Notion (via API)
  - PDF (via converter)

Structured JSON Export

{
  "prd": {
    "metadata": {
      "title": "Smart Search with AI Suggestions",
      "version": "0.1",
      "status": "draft",
      "created": "2026-01-24",
      "author": "PM Name"
    },
    "sections": {
      "overview": {...},
      "goals": {...},
      "requirements": {...},
      "design": {...},
      "launch": {...}
    }
  },
  "validation": {
    "completeness": 0.85,
    "missingSections": ["competitive_analysis"],
    "warnings": ["No wireframes linked"]
  }
}

Best Practices

Start with Problem: Always anchor PRD in user/business problem
Be Specific on Metrics: Vague goals lead to vague outcomes
Include Non-Goals: Explicitly state what you won't do
Link to Research: Reference supporting data
Define Success Clearly: How will you know this worked?
Plan for Failure: Include rollback and contingency plans

Common Pitfalls

Solution-First Thinking: Start with problem, not solution
Missing Metrics: Every PRD needs measurable goals
Scope Creep: Use non-goals to prevent creep
Vague Requirements: Be specific and testable
Missing Dependencies: Identify blockers early

prd-generator