Planner Agent
Researches best practices, analyzes your codebase, and generates step-by-step implementation plans with code examples, timelines, and rollback procedures.
When to Use
- Before major features - Break down complex work into clear steps
- Technical decisions - Evaluate multiple approaches with pros/cons
- Large refactors - Map dependencies and impact before touching code
- CI/CD failures - Analyze logs and create systematic fix plans
Key Capabilities
| Capability | What It Does |
|---|---|
| Research | Searches industry standards, official docs, proven solutions |
| Analysis | Reads codebase, evaluates dependencies, identifies integration points |
| Planning | Breaks work into tasks, lists file changes, estimates timeline |
| Risk Management | Includes rollback plans, security checks, performance considerations |
Common Use Cases
Feature Planning
- Who: Backend dev adding real-time notifications
- Prompt:
/plan [add WebSocket notifications with Socket.io and Redis] - Output: Plan with setup steps, auth integration, database schema, test strategy
Architecture Review
- Who: Tech lead evaluating database migration
- Prompt:
/plan [migrate from MongoDB to PostgreSQL] - Output: Migration strategy, data transformation steps, zero-downtime approach
Bug Investigation
- Who: Developer fixing complex race condition
- Prompt:
/plan:hard [fix checkout race condition causing double charges] - Output: Root cause analysis, reproduction steps, fix plan with test cases
Optimization Planning
- Who: DevOps engineer improving performance
- Prompt:
/plan [optimize API response time from 2s to 200ms] - Output: Profiling strategy, bottleneck analysis, optimization steps with benchmarks
CRO Implementation
- Who: Growth engineer improving conversion
- Prompt:
/plan:cro [improve checkout abandonment rate] - Output: A/B test plan, UX improvements, tracking implementation
What You Get
Every plan includes:
# Implementation Plan: [Feature Name]
## Approach
Why this solution + alternatives considered
## Steps
1. Install Dependencies (5 min)
- Commands to run
2. Core Implementation (20 min)
- Files to create: src/feature/service.ts
- Files to modify: src/server.ts
- Code snippets showing structure
3. Integration (15 min)
- Where to hook into existing code
4. Testing (20 min)
- Test files to create
- Coverage requirements
## Timeline
Total: 1 hour
## Rollback Plan
Step-by-step recovery if issues occur
## Security Checklist
- Auth validation
- Input sanitization
- Rate limiting
## Next Steps
Ready to implement? Run: /cook @plans/your-plan.md
Plans saved to: plans/[feature-name]-YYYYMMDD-HHMMSS.md
Pro Tips
Review before coding: Plans catch design flaws before you write code. Check the approach section first.
Use plans as specs: Reference the plan during code review. “Did we implement step 3?” ensures nothing is missed.
Multiple approaches: Use /plan:two [description] to generate two different solutions with trade-off comparison.
Link from issues: Save plans to git, link from GitHub issues. Future devs understand why decisions were made.
Update after implementation: Mark completed steps, note deviations. Plans become living documentation.
Example Commands
# Basic planning
/plan [add OAuth2 authentication]
# Compare approaches
/plan:two [use Redis vs PostgreSQL for caching]
# Fix planning
/plan:hard [memory leak in job processor]
# CRO planning
/plan:cro [improve signup conversion]
# CI/CD fix
/plan:ci [github-actions-url]
# Execute plan
/cook @plans/oauth2-auth-20241020.md
Integration with Commands
| Phase | Command | Purpose |
|---|---|---|
| Research | /plan | Generate implementation plan |
| Review | cat plans/latest.md | Review plan before coding |
| Execute | /cook @plans/plan.md | Implement following plan |
| Test | /test | Validate implementation |
| Document | Update plan with actuals | Create decision record |
Related Agents
- Fullstack Developer - Executes implementation plans
- Researcher Agent - Deep dives on specific topics
- Scout - Explores codebase for context
Key Takeaway
Planning prevents waste. 10 minutes of research and analysis saves hours of refactoring when you discover you picked the wrong approach halfway through implementation.