Writing Plans
Overview
Write comprehensive implementation plans assuming the engineer has zero context for our codebase and questionable taste. Document everything they need to know: which files to touch for each task, code, testing, docs they might need to check, how to test it. Give them the whole plan as bite-sized tasks. DRY. YAGNI. TDD. Frequent commits.
Assume they are a skilled developer, but know almost nothing about our toolset or problem domain. Assume they don't know good test design very well.
Announce at start: "I'm using the writing-plans skill to create the implementation plan."
Context: This should be run in a dedicated worktree (created by brainstorming skill).
Save plans to: docs/plans/NNNN.YYYY-MM-DD-<feature-name>.md (NNNN: 해당 폴더 내 최대 번호 + 1, 없으면 0001)
Bite-Sized Task Granularity
Each step is one action (2-5 minutes):
- "Write the failing test" - step
- "Run it to make sure it fails" - step
- "Implement the minimal code to make the test pass" - step
- "Run the tests and make sure they pass" - step
- "Commit" - step
Plan Document Header
Every plan MUST start with this header:
# [Feature Name] Implementation Plan
> **For Claude:** REQUIRED SUB-SKILL: Use superpowers:executing-plans to implement this plan task-by-task.
**Goal:** [One sentence describing what this builds]
**Architecture:** [2-3 sentences about approach]
**Tech Stack:** [Key technologies/libraries]
**Constraints:**
- Technical: [기술 스택/인프라/호환성 제약]
- Resource: [시간/인원/예산/환경 제약]
- Business: [법적/규제/계약/정책 제약]
- Scope: [하지 않을 것, 범위 외]
**Business Rules:**
| Rule | Rationale (근거) | Source (출처) |
|------|-------------------|---------------|
| [도메인 규칙] | [왜 이렇게 결정했는가] | [출처] |
> 상위 문서(brainstorming design / PRD)에서 Constraints와 Business Rules를 인계받아 기재한다. 상위 문서가 없으면 사용자에게 직접 확인한다.
---
Task Structure
### Task N: [Component Name]
**Files:**
- Create: `exact/path/to/file.py`
- Modify: `exact/path/to/existing.py:123-145`
- Test: `tests/exact/path/to/test.py`
**Step 1: Write the failing test**
```python
def test_specific_behavior():
result = function(input)
assert result == expected
Step 2: Run test to verify it fails
Run: pytest tests/path/test.py::test_name -v
Expected: FAIL with "function not defined"
Step 3: Write minimal implementation
def function(input):
return expected
Step 4: Run test to verify it passes
Run: pytest tests/path/test.py::test_name -v
Expected: PASS
Step 5: Commit
git add tests/path/test.py src/path/file.py
git commit -m "feat: add specific feature"
## Plan Sanity Check (Devil's Advocate)
플랜을 파일에 저장하기 직전, 아래 3가지를 자가 점검하고 결과를 사용자에게 공유:
1. **순서 공격**: "Task 순서를 뒤집거나 병합하면 더 효율적이지 않은가?"
2. **누락 점검**: "가장 간과하기 쉬운 엣지 케이스 또는 의존성은?"
3. **복잡도 공격**: "이 계획이 과도하게 복잡하지 않은가? Task 수를 절반으로 줄일 수 있는가?"
> 계획 단계에서 발견하면 코스트 1, 구현 중 발견하면 코스트 10.
> 점검 결과 문제가 있으면 수정 후 저장. 문제 없으면 그대로 진행.
## Remember
- Exact file paths always
- Complete code in plan (not "add validation")
- Exact commands with expected output
- Reference relevant skills with @ syntax
- DRY, YAGNI, TDD, frequent commits
---
## Plan Mode Output Capture
Capture TODO lists generated from Plan mode and save them as structured markdown files.
**Triggers:** "플랜 저장해줘", "TODO 저장", "계획 캡처", "save plan", "capture plan"
**Announce at start:** "I'm capturing the Plan mode output to save as structured markdown."
### Storage Mode Selection
Ask user first:
**"How would you like to save the Plan output?**
1. **Task-based** - Each TODO as separate file (enables parallel work)
2. **Step-based** - Single file with sequential checklist (single workflow)
**Please choose."**
### Task-based Storage
Save each TODO as separate file:
**Location:** `docs/plans/<feature>/`
docs/plans/<feature>/ ├── _index.md # Overview + task links ├── task-01-<n>.md ├── task-02-<n>.md └── task-03-<n>.md
**_index.md format:**
```markdown
# [Feature Name] Implementation Plan
**Created:** YYYY-MM-DD HH:mm
**Status:** In Progress
## Overview
[Context analyzed from Plan mode]
## Task List
| # | Task | Status | File |
|---|------|--------|------|
| 1 | [Task name] | ⬜ | [task-01-name.md](./task-01-name.md) |
| 2 | [Task name] | ⬜ | [task-02-name.md](./task-02-name.md) |
## Dependencies
[Task dependency description]
Individual task file format:
# Task 1: [Task Name]
**Status:** ⬜ Pending | 🔄 In Progress | ✅ Complete
**Estimated time:** N min
**Dependencies:** None | After Task N
## Goal
[What this task achieves]
## Work Items
### Step 1: [Step name]
- [ ] Detail 1
- [ ] Detail 2
### Step 2: [Step name]
- [ ] Detail
## Related Files
- `path/to/file.py`
## Completion Criteria
- [ ] Tests pass
- [ ] Code review complete
Step-based Storage
Single file with sequential checklist:
Location: docs/plans/NNNN.YYYY-MM-DD-HHmm-<feature>.md
# [Feature Name] Implementation Plan
**Created:** YYYY-MM-DD HH:mm
**Status:** In Progress
## Overview
[Context analyzed from Plan mode]
---
## Checklist
### Phase 1: [Phase name]
- [ ] Step 1: [Description]
- [ ] Step 2: [Description]
- [ ] Step 3: [Description]
### Phase 2: [Phase name]
- [ ] Step 4: [Description]
- [ ] Step 5: [Description]
---
## Progress Log
| Time | Completed Item | Notes |
|------|----------------|-------|
| | | |
Required Information to Capture
Always extract from Plan mode output:
- Context - Why this work is being done
- TODO list - All work items
- Dependencies - Items with ordering constraints
- Related files - Files to modify/create
- Decisions - Technical decisions made during planning
Post-save Message
✅ Plan saved!
📁 Location: docs/plans/<path>
📋 Tasks: N items
Next steps:
1. Review saved files
2. Modify if needed
3. To start execution, say "execute plan" or "플랜 실행해줘"
Execution Handoff
After saving the plan, offer execution choice:
"Plan complete and saved to docs/plans/<filename>.md. Two execution options:
1. Subagent-Driven (this session) - I dispatch fresh subagent per task, review between tasks, fast iteration
2. Parallel Session (separate) - Open new session with executing-plans, batch execution with checkpoints
Which approach?"
If Subagent-Driven chosen:
- REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development
- Stay in this session
- Fresh subagent per task + code review
If Parallel Session chosen:
- Guide them to open new session in worktree
- REQUIRED SUB-SKILL: New session uses superpowers:executing-plans
Completion
플랜이 docs/plans/에 저장되고 실행 방식(Subagent-Driven / Parallel Session)이 선택되면 완료.
Troubleshooting
Plan tasks are too large (>5 minutes each): Split further. Each step should be one action: write test, run test, implement, commit. If a step has "and", it's two steps. User's Plan mode output is unstructured: Extract TODO items manually. Look for numbered lists, bullet points, or action verbs. Ask user to confirm the extracted task list before saving. Plan becomes outdated during implementation: Plans are living documents. Update the plan file as decisions change. Add a "Changes from original plan" section at the bottom.