# create-plan > Create detailed implementation plans with diagrams and dependency graphs. Use when planning features, refactoring, or multi-step development work before writing code. - Author: Brian Dopson - Repository: dopsonbr/agent-kit - Version: 20251224110637 - Stars: 1 - Forks: 0 - Last Updated: 2026-02-07 - Source: https://github.com/dopsonbr/agent-kit - Web: https://mule.run/skillshub/@@dopsonbr/agent-kit~create-plan:20251224110637 --- --- name: create-plan description: Create detailed implementation plans with diagrams and dependency graphs. Use when planning features, refactoring, or multi-step development work before writing code. license: MIT metadata: author: agent-kit version: "2.1.0" --- # Create Implementation Plan Generate structured, executable implementation plans with visual diagrams and dependency tracking. ## Purpose Transform requirements into actionable plans that: - Stay under 5000 tokens (hard cap) - Include visual diagrams (Mermaid or ASCII) - Show task/subplan dependencies - Are numbered sequentially in `docs/plans/` ## When to Use - Before implementing a new feature - When refactoring existing code - For breaking down complex changes - After brainstorming, before coding ## Critical Constraints ### Token Limit: Hard Cap 5000 - **Optimal:** 2000-3000 tokens (fits well in context with code) - **Maximum:** 5000 tokens (STOP if exceeded) - If plan would exceed 5000 tokens, create subplans instead - Validate with: `bun run scripts/validate-plan.ts ` ### Reference, Never Copy - Reference docs by path only: `per docs/standards/validation.md` - NEVER paste external content into the plan - Only include directly relevant references ## Workflow ### Phase 1: Gather Requirements **Ask clarifying questions before planning:** 1. **Scope**: What exactly should this feature do? 2. **Boundaries**: What's explicitly out of scope? 3. **Dependencies**: What existing code/systems does this touch? 4. **Constraints**: Any technical requirements or limitations? 5. **Success Criteria**: How do we know it's done? Do NOT proceed until you have clear answers. ### Phase 2: Explore Codebase Use the Explore subagent to understand: - Existing modules and patterns - Files that will need modification - Dependencies and integration points - Documentation that needs updates ### Phase 3: Design Architecture Create diagrams to illustrate: - Component relationships - Data/control flow - State transitions - Interactions and sequences Use Mermaid (preferred) or ASCII. Choose the diagram type that best fits: **Flowchart** - Component relationships, data flow: ```mermaid graph TD A[Component A] --> B[Component B] B --> C[Component C] ``` **Sequence** - API calls, interactions between systems: ```mermaid sequenceDiagram Client->>+API: POST /users API->>+DB: INSERT user DB-->>-API: user record API-->>-Client: 201 Created ``` **State** - Lifecycle, status transitions: ```mermaid stateDiagram-v2 [*] --> Draft Draft --> Review Review --> Approved Review --> Draft: Changes requested Approved --> [*] ``` **Class** - Object relationships, data models: ```mermaid classDiagram User "1" --> "*" Order Order "1" --> "*" LineItem ``` **ASCII** - When Mermaid isn't rendered: ``` ┌──────────────┐ ┌──────────────┐ │ Component A │ ──────► │ Component B │ └──────────────┘ └──────────────┘ ``` ### Phase 4: Plan Tasks Break work into phases and tasks: - Each task should be 2-15 minutes - Include file paths (CREATE/MODIFY/DELETE) - Add verification steps - Include commit messages ### Phase 5: Check Size & Split Before writing: 1. Estimate token count (~4 chars = 1 token) 2. If > 3000 tokens, consider splitting 3. If > 5000 tokens, MUST create subplans **Subplan Strategy:** ``` 030_feature-initiative.md (parent overview) ├── 030A_frontend-component.md (subplan A) ├── 030B_backend-api.md (subplan B) └── 030C_integration.md (subplan C) ``` ### Phase 6: Define Testing Strategy Every plan MUST include a testing strategy with: 1. **Automated Tests** (required) - Prefer E2E tests that validate the full feature flow - Include integration tests for component interactions - Add unit tests for complex logic 2. **Manual Validation** (required) - Step-by-step instructions for human verification - What to look for, what to click, what to expect - Screenshots or recordings if helpful ```markdown ## Testing Strategy ### Automated Tests | Type | What It Tests | Command | |------|---------------|---------| | E2E | Full user flow | `bun test:e2e tests/e2e/feature.spec.ts` | | Integration | API + DB | `bun test tests/integration/` | | Unit | Core logic | `bun test tests/unit/` | ### Manual Validation 1. Navigate to /settings 2. Click "Preferences" tab 3. Toggle dark mode 4. **Expected:** Theme changes immediately 5. Refresh page 6. **Expected:** Setting persists ``` ### Phase 7: Create Dependency Graph Every plan MUST end with a dependency graph showing: - Which tasks/subplans can run in parallel - Which must be sequential - External prerequisites ```mermaid graph TD subgraph "Prerequisites" P1[Prior Plan 029] end subgraph "This Plan" A[Phase 1: Setup] --> B[Phase 2: Core] A --> C[Phase 3: Tests] B --> D[Phase 4: Integration] C --> D end P1 --> A ``` ### Phase 8: Write Plan Write to: `docs/plans/{NNNN}_{FEATURE_NAME}.md` Use the template from [assets/plan-template.md](assets/plan-template.md). ### Phase 9: Validate Run the validation script: ```bash bun run content/skills/plan-create/scripts/validate-plan.ts docs/plans/{plan-file}.md ``` ### Phase 10: Commit Plan After validation passes, commit the plan and any related artifacts: ```bash git add docs/plans/{plan-file}.md git add docs/plans/{plan-file}*.md # Include any subplans git commit -m "docs(plan): add {NNNN}_{FEATURE_NAME} - {Brief description of what the plan covers} - Includes dependency graph and testing strategy 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude " ``` Always commit immediately after validation to ensure plans are tracked in version control. ### Phase 11: Offer Review After committing, ask: > "Would you like me to run `/review-plan` to validate this plan before implementation?" The review-plan skill will: - Check for completeness and clarity - Validate dependencies are correct - Ensure testing strategy is adequate - Identify potential issues or gaps ### Phase 12: Offer Worktree Setup After plan is complete, ask: > "Would you like me to create a git worktree for implementing this plan?" Worktrees are created in `.worktrees/` named after the plan: ```bash git worktree add .worktrees/0042_user-preferences -b feature/0042_user-preferences ``` See [references/worktree-setup.md](references/worktree-setup.md) for details. ## Plan Numbering 1. Check `docs/plans/` for the highest existing plan number 2. Use the next sequential number 3. Format: `{NNNN}_{FEATURE_NAME}.md` (4 digits, zero-padded) Examples: - `0001_user-authentication.md` - `0042_api-refactor.md` - `0100_database-migration.md` ## Output Location All plans go to: `docs/plans/` ``` docs/plans/ ├── 0001_initial-setup.md ├── 0002_user-auth.md ├── 0003A_frontend-forms.md ├── 0003B_backend-validation.md └── 0003C_integration-tests.md ``` ## Plan Structure See [assets/plan-template.md](assets/plan-template.md) for the full template. Key sections: 1. **Header** - Status, overview, goals 2. **References** - Only directly relevant docs 3. **Architecture** - Mermaid/ASCII diagrams 4. **Phases** - Numbered tasks with prereqs 5. **Files Summary** - CREATE/MODIFY/DELETE table 6. **Testing Strategy** - Automated tests + manual validation (required) 7. **Dependency Graph** - Required at end 8. **Checklist** - Completion tracking ## Examples See [references/examples.md](references/examples.md) for complete plan examples. ## Validation Run validation before finalizing: - Token count under 5000 - Testing strategy present (automated + manual) - Dependency graph present - Proper numbering format ## Related Skills - `brainstorm` - Use before planning to explore ideas - `review-plan` - Review a plan before execution - `implement-plan` - Execute a plan task by task