2.3 KiB
2.3 KiB
| name | description |
|---|---|
| planning-work | Creates comprehensive implementation plans with bite-sized tasks. Use when you have clear requirements and need a step-by-step guide for execution. |
Planning Work (Writing Plans)
When to use this skill
- After requirements are clear (e.g., post-brainstorming).
- Before modifying any code for a multi-step task.
- When you need to generate a
docs/plans/YYYY-MM-DD-feature.mdfile.
Workflow
- Analyze Context: Understand codebase and questionable taste.
- Create Header: Start plan with the required header and goal.
- Define Architecture: Summarize approach in 2-3 sentences.
- Break Down Tasks: Create bite-sized (2-5 min) tasks.
- Write Task Steps: For each task, define files, tests, and step-by-step TDD instructions.
- Save Plan: Save to
docs/plans/YYYY-MM-DD-<feature-name>.md. - Confirm: Present options for execution.
Instructions
1. Plan Philosophy
- Assume the executor has zero context.
- Bite-Sized: Each task should be 2-5 minutes of work.
- Micro-Steps: "Write failing test" -> "Verify failure" -> "Implement minimal code" -> "Verify pass" -> "Commit".
- DRY, YAGNI, TDD.
2. Plan Header Template
Every plan MUST start with:
# [Feature Name] Implementation Plan
**Goal:** [One sentence describing what this builds]
**Architecture:** [2-3 sentences about approach]
**Tech Stack:** [Key technologies/libraries]
---
3. Task Structure Template
### 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**
[Code Block]
**Step 2: Run test (Expect Fail)**
`command to run test`
**Step 3: Write minimal implementation**
[Code Block]
**Step 4: Run test (Expect Pass)**
`command to run test`
**Step 5: Commit**
`git commit -m "feat: ..."`
4. Execution Handoff
After saving the plan, offer:
- Subagent-Driven: You iterate task-by-task in this session.
- Parallel Session: User opens a new session/window to execute.
Key Principles
- Exact file paths always.
- Complete code snippets (no "add validation logic", write the logic).
- Exact commands for running tests.
- Reference other skills if needed.