| name | planning-and-task-breakdown |
| description | Decompose approved designs into bite-sized implementation tasks with exact file paths, verification steps, and dependency ordering. Use after brainstorming/spec approval, before implementation begins. Each task should take 2-10 minutes for an agent to complete.
|
| version | 1.0.0 |
| permissions | ["fs.read","fs.write"] |
| triggers | [{"context":"design has been approved"},{"context":"spec has been approved"},{"context":"ready to implement"},{"command":"/plan"}] |
| platforms | ["claude-code","cursor","gemini-cli","copilot","codex","opencode","windsurf","kiro"] |
| dependencies | ["brainstorming@^1.0.0"] |
| model_variants | {"claude":{"enforcement_style":"human-partner collaborative","task_phrasing":"Let's break this down into small, focused tasks that we can implement and verify one at a time."},"gemini":{"enforcement_style":"structured-checklist","task_phrasing":"TASK TEMPLATE: ☐ File paths ☐ What to implement ☐ Test expectations ☐ Verification command ☐ Dependencies"},"gpt":{"enforcement_style":"directive-imperative","task_phrasing":"Each task MUST include: exact file paths, implementation scope, test expectations, verification command."}} |
| author | codehands-core |
| signed | true |
| tier | 1 |
Overview
Good plans produce good code. This skill transforms approved designs into a sequence of small, independent, verifiable implementation tasks. Each task is scoped so an agent can complete it in a single focused cycle — read the task, implement, test, commit.
When to Use
- After
codehands:brainstorming design approval
- After
codehands:spec-driven-development spec sign-off
- Before any implementation begins
- When an existing plan needs re-planning due to scope changes
- NOT during implementation (if you need to re-plan mid-implementation, pause and re-enter this skill)
Process
1. Read the Spec Completely
Read the approved design/spec document end-to-end. Do not start decomposing until you've read it all. You're looking for:
- Every requirement that requires code changes
- Dependencies between requirements
- Shared infrastructure needed by multiple features
- Testing requirements
2. Identify Implementation Order
Dependencies dictate order. Build in this sequence:
- Shared infrastructure — Types, interfaces, utilities used by multiple components
- Data layer — Database schemas, models, data access
- Business logic — Core functions, services, algorithms
- Integration layer — API endpoints, event handlers, coordinators
- Presentation — UI components, CLI output, response formatting
- Polish — Error messages, logging, documentation
3. Write Task Definitions
Each task MUST include ALL of these fields:
### Task N: [Brief descriptive title]
**Files:**
- `src/models/user.ts` (CREATE)
- `src/services/auth.ts` (MODIFY, lines 45-60)
- `tests/models/user.test.ts` (CREATE)
**Implement:**
- [Specific behavior to implement]
- [Exactly what to build — no ambiguity]
- [Include code patterns to follow if relevant]
**Test Expectations:**
- [What the test should verify]
- [Expected inputs and outputs]
**Verification:**
```bash
npm test -- --grep "User model"
Depends On: Task 1, Task 3
Estimated Scope: [Small/Medium/Large]
### 4. Size Each Task Correctly
| Size | Lines of Code | Time | When to Use |
|---|---|---|---|
| **Small** | < 50 LOC | 2-5 min | Single function, simple test |
| **Medium** | 50-150 LOC | 5-10 min | Component + tests, API endpoint |
| **Large** | 150+ LOC | 10-15 min | Maximum size — consider splitting |
> **If a task would take > 15 minutes or touch > 3 files, split it.**
### 5. Verify the Plan
Before presenting to the user:
- [ ] Every spec requirement is covered by at least one task
- [ ] Dependencies are acyclic (no circular dependencies)
- [ ] Each task can be verified independently
- [ ] The first task can be started without any other task completing
- [ ] Shared infrastructure tasks come before tasks that depend on them
- [ ] Tests are included IN each task (not "add tests later")
### 6. Save and Get Approval
Save the plan to `docs/plans/YYYY-MM-DD-<topic>-plan.md` and present to the user:
> "Implementation plan written to `<path>`. It has N tasks with an estimated scope of [Small/Medium/Large]. Review and let me know if you'd like any changes before I start."
## Plan Format
```markdown
# Implementation Plan: [Feature Name]
**Spec:** `docs/specs/YYYY-MM-DD-<topic>-spec.md`
**Created:** YYYY-MM-DD
**Tasks:** N total | N small | N medium | N large
**Estimated Total Scope:** [Sum of task sizes]
## Dependency Graph
Task 1 → Task 3 → Task 5
Task 2 → Task 3
Task 4 (independent)
## Tasks
### Task 1: [Title]
[Full task definition as specified above]
### Task 2: [Title]
...
Common Rationalizations
| Rationalization | Reality |
|---|
| "I can keep the plan in my head" | Context windows are finite and unreliable. Written plans survive context switches, session restarts, and hand-offs. Always write it down. |
| "The spec is clear enough to implement directly" | Specs describe WHAT. Plans describe HOW and IN WHAT ORDER. Even clear specs need decomposition into verifiable steps. |
| "This is only 2-3 tasks, I don't need formal planning" | Even 2-3 tasks benefit from explicit file paths, test expectations, and verification commands. The format takes 2 minutes to write and saves 20 minutes of confusion. |
| "I'll figure out the order as I go" | Implementing without dependency ordering causes rework. Building a component before its dependency is ready means guessing at interfaces. |
| "Tests can be a separate task at the end" | Tests in each task, not at the end. Deferring tests creates a "testing debt" that never gets paid. TDD means the test IS the task. |
Red Flags
- Tasks without file paths (where will the code go?)
- Tasks without verification commands (how will you know it works?)
- Tasks that touch 5+ files (too big — split it)
- "Add tests" as the last task
- No dependency ordering
- Tasks that duplicate spec content verbatim without decomposition
Verification
See Also
codehands:brainstorming — Design phase that precedes planning
codehands:spec-driven-development — Spec that drives the plan
codehands:multi-agent-orchestration — Execute the plan with subagents
codehands:executing-plans — Execute the plan in parallel sessions
codehands:context-engineering — Manage context during planning