Building Software Like a Master Architect
Imagine you want to build a house. You wouldn't just grab a hammer and start hitting wood, right? You'd hire an architect to draw plans, a structural engineer to figure out the foundation, and skilled builders to follow the blueprints.
The PRP Framework works exactly the same way for building software.
Instead of randomly writing code and hoping it works, you use specialized "AI construction workers" who follow detailed blueprints to build your software features correctly the first time.
Think of the PRP commands as different specialists in a construction company:
- What they do: Take your vague idea ("I want a nice house") and create detailed architectural plans with blueprints, materials list, and construction phases
- When to use: When you have a rough idea but need comprehensive planning
- Example: "I want users to be able to like posts" β Complete feature specification with diagrams
- What they do: Create detailed specifications for how different parts connect (like how the foundation connects to the walls)
- When to use: After the architect creates plans, when you need frontend and backend to work together
- Example: Defines exactly how the "like button" talks to the database
- What they do: Create incredibly detailed construction manuals with every nail, screw, and measurement specified
- When to use: When you need comprehensive implementation instructions
- Example: Step-by-step instructions for building the entire like button feature
- What they do: Plan how to change existing buildings (like converting a garage into a bedroom)
- When to use: When you need to modify existing code rather than build something new
- Example: Changing from simple authentication to OAuth2
- What they do: Create focused daily work orders with quality checks
- When to use: For small, specific tasks that need careful execution
- Example: "Add a dark mode toggle to the settings page"
- What they do: Actually build the entire feature following the detailed plans
- When to use: After you have comprehensive building plans
- Example: Takes the like button plans and builds the working feature
- What they do: Execute renovation plans, carefully transforming existing structures
- When to use: After you have renovation specifications
- Example: Actually migrates your authentication system
- What they do: Handle focused tasks with surgical precision
- When to use: After you have specific task lists
- Example: Adds the dark mode toggle exactly as specified
- What they do: Create rapid action plans for urgent situations
- When to use: Hackathons, urgent fixes, or when you need to move fast
- Example: "Build a social dashboard in 2 days"
This is the KEY part that makes everything work!
Just like real construction, each specialist builds on the work of the previous one:
ποΈ Master Architect creates building plans
β (Plans get passed to...)
π Structural Engineer creates connection specs
β (Both plans and specs get passed to...)
π Detail Architect creates construction manual
β (Manual gets passed to...)
π¨ Master Builder builds the actual feature
IMPORTANT: Each step uses the output from previous steps - they're not isolated tools!
Step 1: Get the Big Picture
/prp-planning-create "user profile page with avatar upload and bio editing"- Creates:
PRPs/user-profile-prd.md(architectural plans) - Contains: Complete feature specification, user flows, technical architecture
Step 2: Define How Parts Connect
/api-contract-define "create API endpoints for the user profile feature described in PRPs/user-profile-prd.md"- Creates:
PRPs/contracts/user-profile-api-contract.md(connection specs) - Contains: Exact API endpoints, data structures, error handling
Step 3: Create Detailed Instructions
/prp-base-create "implement user profile feature using PRPs/user-profile-prd.md architecture and PRPs/contracts/user-profile-api-contract.md specifications"- Creates:
PRPs/user-profile-implementation.md(construction manual) - Contains: Step-by-step implementation with all context needed
Step 4: Build It
/prp-base-execute PRPs/user-profile-implementation.md- Result: Working user profile feature with all tests passing
Step 1: Plan the Changes
/prp-spec-create "migrate user authentication from basic auth to OAuth2 with Google integration"- Creates:
SPEC_PRP/PRPs/oauth2-migration.md(renovation plans) - Contains: Current state, desired state, transformation steps
Step 2: Execute the Changes
/prp-spec-execute SPEC_PRP/PRPs/oauth2-migration.md- Result: Authentication system successfully migrated
Step 1: Define the Task
/prp-task-create "add email validation to the signup form with proper error messages"- Creates:
TASK_PRP/PRPs/email-validation.md(work order) - Contains: Specific tasks with validation and rollback plans
Step 2: Do the Task
/prp-task-execute TASK_PRP/PRPs/email-validation.md- Result: Email validation added with proper testing
Step 1: Create Action Plan
/task-list-init "hackathon project: social media dashboard with posts, likes, and user profiles"- Creates:
PRPs/checklist.md(emergency action plan) - Contains: Prioritized task list with status tracking
Step 2: Execute Manually
- Follow the checklist, checking off items as you complete them
- Use other commands for specific parts if needed
Let's build a social media like button from scratch, step by step:
/prp-planning-create "social media like button with real-time updates and analytics tracking"What happens behind the scenes:
- AI researches existing like button implementations
- Creates user flow diagrams
- Designs database schema for likes
- Plans real-time update architecture
- Specifies analytics requirements
- Creates comprehensive PRD document
Result: PRPs/like-button-prd.md with complete architectural plans
/api-contract-define "create API contracts for the like button feature specified in PRPs/like-button-prd.md, including real-time updates and analytics"What happens behind the scenes:
- Reads the architectural plans from Step 1
- Defines specific API endpoints:
POST /api/posts/{id}/likeDELETE /api/posts/{id}/likeGET /api/posts/{id}/likes/analytics
- Creates TypeScript interfaces
- Specifies WebSocket events for real-time updates
- Defines error responses
Result: PRPs/contracts/like-button-api-contract.md with exact technical specifications
/prp-base-create "implement like button feature using architecture from PRPs/like-button-prd.md and API specifications from PRPs/contracts/like-button-api-contract.md"What happens behind the scenes:
- Reads BOTH the architectural plans AND the API contracts
- Researches the existing codebase for patterns
- Creates step-by-step implementation plan
- Includes all necessary context (documentation links, code examples)
- Defines 4-level quality checking process
- Creates comprehensive construction manual
Result: PRPs/like-button-implementation.md with everything needed to build it
/prp-base-execute PRPs/like-button-implementation.mdWhat happens behind the scenes:
- Planning Phase: Creates detailed todo list using TodoWrite
- Foundation (Level 1):
- Creates database migration for likes table
- Sets up TypeScript interfaces
- Runs syntax checking:
ruff check --fix && mypy .
- Structure (Level 2):
- Implements API endpoints following the contract
- Creates React component with proper state management
- Adds comprehensive unit tests
- Runs tests:
uv run pytest tests/ -v
- Systems (Level 3):
- Integrates WebSocket for real-time updates
- Connects frontend to backend APIs
- Tests complete user workflows
- Runs integration tests
- Final Inspection (Level 4):
- Load tests with multiple concurrent users
- Tests edge cases and error scenarios
- Verifies analytics accuracy
Result: Fully working like button with real-time updates, analytics, and all tests passing!
/prp-task-create "add animated heart floating effect when users like posts for better visual feedback"What happens: Creates focused task list for the animation enhancement
/prp-task-execute TASK_PRP/PRPs/heart-animation.mdResult: Beautiful floating heart animation added to the like button
Every feature goes through 4 levels of quality control, just like building inspections:
ruff check --fix && mypy .Like checking if the foundation is level and meets building codes
uv run pytest tests/ -v Like testing if each room functions correctly (lights work, doors open, etc.)
curl -X POST http://localhost:8000/api/posts/123/likeLike testing if plumbing, electrical, and heating all work together
Like having a family actually live in the house to find any remaining issues
- You have a vague idea that needs to become a concrete plan
- You're starting a new major feature
- You need to research and understand the problem space
- You want comprehensive documentation with diagrams
- You have architectural plans and need technical specifications
- Frontend and backend teams need to coordinate
- You need exact API endpoint definitions
- You want to prevent integration issues
- You need comprehensive implementation instructions
- You're building something completely new
- You want all the context and examples included
- You need the full 4-level validation process
- You need to modify existing code
- You're doing a migration or refactoring
- You need to document current vs. desired state
- You want rollback plans for safety
- You have a small, focused change to make
- You need surgical precision on specific code
- You want immediate validation after each step
- The change affects only a few files
- You have the corresponding plan/spec/task document
- You're ready to actually build/change the code
- You want systematic, validated implementation
- You trust the AI to follow the detailed instructions
- You're in a hackathon or time crunch
- You need a quick overview of what needs to be done
- You want to track progress manually
- You're doing rapid prototyping
# WRONG - each command creates isolated work
/api-contract-define "user authentication"
/prp-base-create "user authentication" # RIGHT - each command builds on the previous
/prp-planning-create "user authentication with social login"
/api-contract-define "create API for authentication described in PRPs/user-auth-prd.md"
/prp-base-create "implement authentication using PRPs/user-auth-prd.md and PRPs/contracts/user-auth-api-contract.md"# WRONG - jumping straight to implementation
/prp-base-create "some complicated feature"# RIGHT - plan first, then implement
/prp-planning-create "some complicated feature"
/prp-base-create "implement feature using PRPs/complicated-feature-prd.md"# WRONG - vague and unhelpful
/prp-base-create "make the app better"# RIGHT - specific and actionable
/prp-base-create "add user profile editing with avatar upload, bio editing, and email preferences"For any non-trivial feature, start with /prp-planning-create. The time you spend planning saves hours of implementation confusion.
Always reference previous outputs in your next command. If you created PRPs/user-auth-prd.md, reference it in your next command.
Instead of "add authentication," say "add OAuth2 authentication with Google and GitHub integration, including user profile sync and role-based permissions."
Let each command run through all 4 quality levels. The validation catches issues before they become problems.
Plan the foundation before building the walls. Design the overall structure before implementing individual components.
When creating implementation PRPs, reference existing code patterns, documentation URLs, and similar implementations in your codebase.
| Command | Purpose | Input | Output |
|---|---|---|---|
/prp-planning-create |
Master planning | Rough idea | Comprehensive PRD with diagrams |
/api-contract-define |
Technical specs | Feature + PRD reference | API contracts and interfaces |
/prp-base-create |
Implementation manual | Feature + all references | Complete construction guide |
/prp-base-execute |
Build new feature | Implementation PRP path | Working feature |
/prp-spec-create |
Renovation plans | Change requirements | Currentβdesired transformation |
/prp-spec-execute |
Execute changes | Spec PRP path | Modified code |
/prp-task-create |
Focused work orders | Specific task | Detailed task list |
/prp-task-execute |
Do specific tasks | Task PRP path | Completed task |
/task-list-init |
Emergency planning | Urgent project | Trackable checklist |
-
Pick something simple like "add a footer to the website with copyright and links"
-
Plan it:
/prp-planning-create "website footer with copyright, privacy policy link, and contact link" -
Create implementation guide:
/prp-base-create "implement footer using architecture from PRPs/website-footer-prd.md" -
Build it:
/prp-base-execute PRPs/website-footer-implementation.md
-
Celebrate! π You just used AI to build a feature with proper planning, documentation, and quality control!
- What they do: Prime Claude with essential project knowledge by reading key files
- When to use: Start of every conversation, after major changes
- Example:
/prime-core - Result: Claude understands your project structure and can work effectively
- What they do: Generate comprehensive onboarding documentation for new developers
- When to use: New team members join, documenting undocumented projects
- Example:
/onboarding - Result: Creates ONBOARDING.md and QUICKSTART.md with complete setup guides
- What they do: Analyze changes and create properly formatted commit messages
- When to use: Before committing code, when following commit conventions
- Example:
/commit "fixing authentication bug" - Result: Properly formatted commit with conventional commit style
- What they do: Create a new git branch following team conventions
- When to use: Starting new feature work, creating bugfix branches
- Example:
/new-dev-branch feature/user-profile - Result: New branch created with proper naming
- What they do: Systematic debugging to find root causes of issues
- When to use: When facing complex bugs, mysterious errors
- Example:
/debug-RCA "users can't login after password reset" - Result: Detailed analysis with root cause and fix recommendations
- What they do: Generate comprehensive PR with description and testing notes
- When to use: After feature completion, ready for code review
- Example:
/create-pr - Result: Well-documented PR ready for review
- What they do: Scan code for refactoring opportunities (Python-focused)
- When to use: Before merging, cleaning technical debt
- Example:
/refactor-simple - Result: Creates refactor_plan.md with prioritized improvements
- What they do: Thorough review checking quality, security, performance
- When to use: Before commits, PR reviews, code audits
- Example:
/review-general src/features/authentication - Result: Detailed review report in PRPs/code_reviews/
- What they do: Review specifically what's staged/unstaged in git
- When to use: Pre-commit review, understanding current changes
- Example:
/review-staged-unstaged - Result: Analysis of current git status with recommendations
- What they do: Intelligently resolve git merge conflicts
- When to use: After merge conflicts, during rebases
- Example:
/conflict-resolver-general - Result: Conflicts resolved with integrity maintained
- What they do: Deep analysis conflict resolution with business logic understanding
- When to use: Complex conflicts, critical business logic conflicts
- Example:
/smart-resolver - Result: Intelligent resolution with detailed documentation
- What they do: Resolve conflicts in specific files only
- When to use: When you only need to resolve certain files
- Example:
/conflict-resolver-specific src/api/auth.ts - Result: Targeted resolution of specified files
- What they do: 15 parallel agents analyze 3 approaches for hackathon challenges
- When to use: Hackathon start, evaluating multiple solutions
- Example:
/hackathon-research "AI-powered code review tool" - Result: Comprehensive analysis with best approach recommendation
- What they do: Generate user stories with acceptance criteria rapidly
- When to use: Sprint planning, feature definition
- Example:
/user-story-rapid "user profile management" - Result: Complete user stories ready for implementation
- What they do: Create multiple PRPs simultaneously using parallel processing
- When to use: Large feature sets, multiple related features
- Example:
/parallel-prp-creation "e-commerce checkout flow" - Result: Multiple coordinated PRPs created efficiently
- What they do: Analyze existing PRPs and optionally execute them
- When to use: Understanding PRP quality, batch execution
- Example:
/prp-analyze-run - Result: Analysis report and execution results
- What they do: Check PRPs against quality standards
- When to use: Before execution, quality assurance
- Example:
/prp-validate PRPs/user-auth.md - Result: Validation report with improvement suggestions
- What they do: Create comprehensive PRPs using parallel research
- When to use: Complex features needing deep research
- Example:
/create-base-prp-parallel "real-time collaboration" - Result: Research-backed comprehensive PRP
- What they do: Create planning PRDs using parallel analysis
- When to use: Initial feature planning, architecture design
- Example:
/create-planning-parallel "mobile app backend" - Result: Comprehensive PRD with multiple perspectives
- What they do: Create rapid implementation PRPs for hackathons
- When to use: Time-constrained development, MVPs
- Example:
/hackathon-prp-parallel "AI chatbot integration" - Result: Streamlined PRP optimized for speed
- What they do: Create PRPs optimized for TypeScript projects
- When to use: TypeScript feature development
- Example:
/TS-create-base-prp "type-safe API client" - Result: TypeScript-focused implementation PRP
- What they do: Execute PRPs with TypeScript-specific validation
- When to use: Implementing TypeScript features
- Example:
/TS-execute-base-prp PRPs/api-client.md - Result: Type-safe implementation with full validation
- What they do: Review with TypeScript best practices focus
- When to use: TypeScript code reviews
- Example:
/TS-review-general src/components - Result: TS-specific review with type safety checks
- What they do: Review staged/unstaged TS changes
- When to use: Pre-commit TS reviews
- Example:
/TS-review-staged-unstaged - Result: TypeScript-focused change analysis
/prime-core- Load project context/onboarding- Understand the project/new-dev-branch- Create feature branch
/prp-planning-create- Comprehensive planning/hackathon-research- Evaluate approaches/user-story-rapid- Quick user stories
/prp-base-createβ/prp-base-execute- New features/prp-spec-createβ/prp-spec-execute- Modifications/prp-task-createβ/prp-task-execute- Small tasks
/review-general- Full code review/refactor-simple- Find improvements/debug-RCA- Fix complex bugs
/smart-commit- Create commits/conflict-resolver-general- Resolve conflicts/create-pr- Create pull requests
/hackathon-research- Fast analysis/parallel-prp-creation- Batch creation/task-list-init- Quick task tracking
Many commands work better together:
/prime-core
/prp-planning-create "feature description"
/api-contract-define "using PRPs/feature-prd.md"
/prp-base-create "using both previous files"
/prp-base-execute PRPs/feature-implementation.md
/smart-commit "feat: add new feature"
/create-prUse parallel commands for speed:
- Research phase:
/hackathon-research - Planning phase:
/create-planning-parallel - Implementation:
/create-base-prp-parallel
Always reference previous outputs:
- Bad:
/prp-base-create "user auth" - Good:
/prp-base-create "implement auth using PRPs/auth-prd.md and PRPs/contracts/auth-api.md"
- Use
/review-staged-unstagedbefore every commit - Run
/refactor-simpleweekly for code health - Apply
/review-generalbefore PRs
- You decide WHAT to build (the vision, requirements, business goals)
- AI figures out HOW to build it (technical implementation, code patterns, testing)
- The PRP system ensures quality (proper planning, validation, documentation)
This isn't about replacing human creativity - it's about amplifying your ideas with systematic, high-quality implementation.
Ready to build something amazing? Start with /prime-core, then /prp-planning-create and watch your ideas come to life! β¨