prp-generator

PRP Generator

Overview

Generates comprehensive Product Requirement Plans (PRPs) that enable AI agents to implement features in a single pass. Combines systematic codebase analysis with external research to create detailed, context-rich implementation blueprints.

When to Use

• User requests a PRP, PRD, or detailed implementation plan
• User asks to "plan out" or "design" a complex feature
• Beginning significant feature development that benefits from structured planning
• User provides a feature description file and asks for implementation guidance

Core Principle

Context is Everything: The implementing agent only receives the PRP content, training data knowledge, codebase access, and WebSearch. Your PRP must be self-contained with all necessary context, specific references, and executable validation gates.

Workflow

Phase 1: Understand the Feature

1. Read the feature request - if a file path is given, read it completely; if verbal, clarify requirements
2. Clarify ambiguities - use AskUserQuestion for unclear requirements, confirm tech stack, verify integration points
3. Identify the core problem being solved and acceptance criteria

Phase 2: Codebase Analysis (Mandatory)

Goal: Understand existing patterns, conventions, and integration points.

Systematically analyse the codebase across five dimensions:

Area	What to Capture
Similar features	File paths, line numbers, code snippets, adaptations needed
Architecture	Directory conventions, component organisation, state management, API patterns
Coding conventions	TypeScript usage, component patterns, styling, import ordering, naming
Test patterns	Framework, file naming, mock strategies, coverage expectations
Configuration	Dependencies, build setup, path aliases, TypeScript settings

For detailed sub-steps, examples, and documentation templates, see references/codebase-analysis-guide.md.

Also refer to references/research_methodology.md for the full research process.

Phase 3: External Research (Mandatory)

Goal: Find best practices, documentation, examples, and gotchas.

Research across four areas:

Area	Key Actions
Library documentation	Find official docs for the SPECIFIC version from package.json; note version-specific gotchas
Implementation examples	Search GitHub, StackOverflow, official examples; prefer recent, production-grade code
Best practices	Search "[technology] best practices [current year]"; check OWASP for security
Performance and security	Bundle size implications, runtime patterns, vulnerabilities, accessibility

Always document exact URLs, versions, and specific sections. See references/research_methodology.md for detailed guidance.

Phase 4: Ultra-Thinking (Critical)

STOP AND THINK DEEPLY BEFORE WRITING THE PRP.

Analyse integration points, implementation ordering, validation strategy, and context completeness. Verify the PRP will enable one-pass implementation without questions.

For the full set of analysis questions and the quality checklist, see references/quality-assessment.md.

Phase 5: Generate the PRP

Use assets/prp_template.md as the base structure. Populate all sections:

1. Metadata - feature name, timeline, confidence score (1-10), date
2. Executive Summary - 2-3 sentences with core value proposition
3. Research Findings - codebase analysis (file:line refs) and external research (URLs, versions)
4. Technical Specification - architecture, components, data models, API endpoints
5. Implementation Blueprint - prerequisites, step-by-step with pseudocode, file changes, error handling, edge cases
6. Testing Strategy - unit, integration, and manual testing approaches
7. Validation Gates - must be EXECUTABLE commands (e.g. npm run test && npm run build)
8. Success Criteria - clear, measurable checklist

Phase 6: Quality Scoring

Score the PRP for one-pass implementation success:

Score	Meaning
9-10	Exceptionally detailed, all context included, clear path, executable gates
7-8	Very good, minor gaps, mostly clear implementation path
5-6	Adequate, some ambiguity, may require clarification
3-4	Incomplete research, missing context, unclear path
1-2	Insufficient for implementation

If score is below 7: Go back and improve the PRP before delivering.

Phase 7: Save and Deliver

1. Save the PRP to PRPs/[feature-name].md (kebab-case, create directory if needed)
2. Deliver summary to user with: brief feature summary, file location, confidence score with rationale, and next steps

Common Pitfalls

Pitfall	Bad	Good
Vague references	"There's a similar component somewhere"	"See UserProfile at src/components/UserProfile.tsx:45-67"
Missing versions	"Use React Query"	"Use @tanstack/react-query v5.28.0"
Non-executable gates	"Run tests and make sure they pass"	npm run test && npm run build
Generic advice	"Follow React best practices"	"Use named exports (see src/components/Button.tsx:1)"
Incomplete research	Skipping codebase analysis	Thoroughly document existing patterns
Missing gotchas	Assuming smooth implementation	Document known issues and edge cases

Example Usage

User: "Create a PRP for adding dark mode support to the application"

1. Clarify: "Should dark mode preference persist across sessions? Should it respect system preferences?"
2. Research codebase for theme-related code
3. Research external resources (dark mode best practices, library options)
4. Ultra-think about implementation approach
5. Generate comprehensive PRP using template
6. Score the PRP
7. Save to PRPs/dark-mode-support.md
8. Deliver summary with confidence score

Resources

Resource	Description
assets/prp_template.md	Base template for all PRPs
references/research_methodology.md	Detailed research guidance and best practices
references/codebase-analysis-guide.md	Detailed codebase analysis sub-steps and examples
references/quality-assessment.md	Ultra-thinking analysis questions and quality checklist

Key Reminders

• Research is mandatory - never skip codebase or external research
• Be specific - always include file paths, line numbers, URLs, versions
• Think deeply - Phase 4 (Ultra-Thinking) is critical for success
• Validate everything - all validation gates must be executable
• Score honestly - if confidence is below 7, improve the PRP
• Context is king - the implementer only has what you put in the PRP