Planning
Transform vague requirements into detailed, executable technical plans through systematic research, codebase analysis, and solution design.
Core Principle
YAGNI, KISS, DRY first. Plans over code. Brutal honesty over politeness.
Before any implementation, a well-researched plan eliminates wasted effort, premature optimization, and architectural regret. This skill doesn’t write code—it creates battle-tested blueprints that junior developers can execute confidently.
When to Use
Always use for:
- New feature implementations
- System architecture decisions
- Technical approach evaluation
- Complex refactoring projects
Especially when:
- Requirements span multiple components
- Trade-offs need formal comparison
- Codebase patterns must be preserved
- Security/performance are critical
The Process
1. Initial Analysis
Read codebase docs, development-rules.md, existing patterns. Understand constraints before proposing solutions.
2. Research Phase
Spawn researcher agents for unknowns. Investigate libraries, patterns, architectural approaches. Skip if reports already provided.
3. Codebase Understanding
Use scout agents to analyze structure, dependencies, conventions. Skip if scout reports provided.
4. Solution Design
Synthesize research + codebase analysis. Design architecture with clear trade-offs. Provide 2-3 options when no clear winner exists.
5. Plan Documentation
Write self-contained plans with:
- Context: Why this problem exists
- Options: Evaluated alternatives with pros/cons
- Recommendation: Chosen approach with rationale
- Phases: Step-by-step implementation breakdown
- Security/Performance: Addressed concerns
- Validation: How to verify success
6. Review & Output
Ensure plan is actionable by junior developers. Include code snippets/pseudocode for clarity. Store in timestamped directory.
Common Use Cases
Implementing OAuth Authentication
Who: Full-stack developer adding social login
"Add Google OAuth to our Next.js app. We use NextAuth but haven't configured providers yet. Users should sign in with Google, store profile in PostgreSQL, and maintain sessions securely."
Migrating to Microservices
Who: Backend architect refactoring monolith
"Plan migration of our user management module from Rails monolith to standalone microservice. Current system handles auth, profiles, and permissions. 50k active users. Zero downtime required."
Real-Time Chat Feature
Who: Product engineer adding messaging
"Design real-time chat for our SaaS dashboard. Users need 1-on-1 and group conversations with message history, typing indicators, and read receipts. Tech stack: React + Node.js. Consider WebSockets vs SSE."
Database Schema Refactoring
Who: Data engineer optimizing queries
"Refactor our e-commerce product catalog schema. Current design has 12 joins for category filtering, causing 3s page loads. 100k products, 5k categories. Must maintain backward compatibility during migration."
CI/CD Pipeline Setup
Who: DevOps engineer automating deployments
"Create CI/CD pipeline for our TypeScript monorepo with 6 apps (3 Next.js, 2 Express APIs, 1 React Native). We use GitHub Actions. Need automated testing, preview deployments, and production releases."
Pro Tips
Not activating? Say: “Use planning skill to create a detailed implementation plan with research and trade-off analysis.”
Active Plan State: Planning skill creates .claude/active-plan to prevent version proliferation. It prompts “Continue with existing plan? [Y/n]” when resuming work.
Directory Structure:
plans/20251208-1430-oauth-implementation/
├── research/ # Researcher agent reports
├── scout/ # Codebase analysis reports
├── reports/ # General findings
├── plan.md # Main plan document
└── phase-01-setup.md # Implementation phases
Report Location: All agents write to {active-plan}/reports/ using format {agent}-{YYMMDD}-{slug}.md. Fallback: plans/reports/
Quality Gates:
- Thorough: Consider edge cases, security, performance
- Specific: Concrete steps, not vague guidance
- Maintainable: Align with codebase patterns
- Actionable: Junior devs can execute without guessing
Related Skills
- Sequential Thinking - Multi-step analysis
- Problem Solving - Complexity resolution
- Research - External investigation
Key Takeaway
Planning creates comprehensive, executable blueprints through research, codebase analysis, and trade-off evaluation. Plans are self-contained documents stored in timestamped directories with active plan tracking to prevent version chaos.