writing-plans
Write implementation plans before coding. Triggers on: 'write plan', 'implementation plan', 'plan this'.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Write implementation plans before coding. Triggers on: 'write plan', 'implementation plan', 'plan this'.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Evaluate architectural decisions, propose ADRs. Triggers on: 'architecture review', 'ADR', 'design review'.
Perform structured code reviews. Triggers on: 'code review', 'review code', 'review PR'.
Create GitHub pull requests. Triggers on: 'create PR', 'pull request', 'submit PR'.
Evaluate strategic direction changes before execution. Triggers on: 'direction change', 'strategic decision', 'pivot'.
Triage and organize GitHub issues. Triggers on: 'triage', 'triage issues', 'organize backlog'.
Create new custom agent definition. Triggers on: 'create agent', 'new agent'.
| name | writing-plans |
| description | Write implementation plans before coding. Triggers on: 'write plan', 'implementation plan', 'plan this'. |
Create an implementation plan before touching code. Plans assume the executing engineer has zero codebase context.
Save to docs/plans/YYYY-MM-DD-<feature-name>.md:
# Plan: [Feature Name]
**Issue**: #N
**Date**: YYYY-MM-DD
**Author**: Copilot
## Goal
[One sentence: what does this accomplish?]
## Affected Files
- `path/to/file1.py` — [what changes]
- `path/to/file2.py` — [what changes]
- `tests/test_file1.py` — [new test file]
## Tasks
### Task 1: [Description]
**Files**: `path/to/file.py`
**Acceptance**: [What "done" looks like]
Steps:
1. Write failing test for [behavior]
2. Run test — expect RED
3. Implement [minimal code]
4. Run test — expect GREEN
5. Commit: "test: add [test name]" / "feat: implement [feature]"
### Task 2: [Description]
...
## Dependencies
- Task 2 depends on Task 1 (uses types defined there)
- Tasks 3 and 4 are independent (can run in parallel)
## Risks
- [Risk 1 — mitigation]
Each task should take 2-5 minutes to execute:
Every task that changes behavior follows Red-Green-Refactor:
Each task must include:
Before executing: