# Plan With Agent

> アイデアをPlans.mdに落とし込む。夢を計画に変える魔法使い。Use when user mentions '/plan-with-agent', create a plan, organize tasks, feature planning, or TDD planning. Do NOT load for: implementation, reviews, or setup.

- Skill: `majiayu000/plan-with-agent-2` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add majiayu000/plan-with-agent-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/plan-with-agent-2/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/majiayu000/plan-with-agent-2

---


# Planning Skill

Organizes ideas and requirements, converting them into executable tasks in Plans.md.

## Quick Reference

- "**Create a plan**" → this skill
- "**Turn what we talked about into a plan**" → extract from conversation
- "**Want to organize what to build**" → start with hearing
- "**Plan with TDD**" → force TDD adoption
- "**CI/benchmark**" → `--ci` mode

## Deliverables

- **Plans.md** - Task list executable with `/work` (required)
- **Feature priority matrix** - Required/Recommended/Optional classification

## Usage

```bash
/plan-with-agent       # Interactive planning
/plan-with-agent --ci  # CI mode (non-interactive)
```

## Mode-specific Usage

| Mode | Command | Description |
|------|---------|-------------|
| **Solo mode** | `/plan-with-agent` | Claude Code alone: plan → execute → review |
| **2-agent mode** | `/plan-with-cc` (Cursor) | Plan with Cursor → Execute with Claude Code |

## Feature Details

| Feature | Reference |
|---------|-----------|
| **Execution Flow** | See [references/execution-flow.md](references/execution-flow.md) |
| **TDD Adoption** | See [references/tdd-adoption.md](references/tdd-adoption.md) |
| **Priority Matrix** | See [references/priority-matrix.md](references/priority-matrix.md) |

## Execution Flow Overview

1. **Step 0**: Check conversation context (extract from previous conversation or start fresh)
2. **Step 1**: Hearing what to build
3. **Step 2**: Increase resolution (max 3 questions)
4. **Step 3**: Technical research (WebSearch)
5. **Step 4**: Extract feature list
6. **Step 5**: Create priority matrix (Required/Recommended/Optional)
7. **Step 5.5**: TDD adoption judgment and test design
8. **Step 6**: Effort estimation (reference)
9. **Step 7**: Generate Plans.md with quality markers
10. **Step 8**: Guide next actions

## Auto-invoke Skills

| Skill | When to Call |
|-------|--------------|
| `setup` | **Call first** (executes adaptive setup) |
| `vibecoder-guide` | When user is non-technical |

## Quality Markers (Auto-assigned)

| Task Content | Marker | Effect |
|--------------|--------|--------|
| Auth/login feature | `[feature:security]` | Security checklist |
| UI component | `[feature:a11y]` | a11y check |
| Business logic | `[feature:tdd]` | TDD recommended |
| API endpoint | `[feature:security]` | Input validation |
| Bug fix | `[bugfix:reproduce-first]` | Reproduction test first |

## Next Actions

After planning:
- Start implementation with `/work`
- Or say "start from Phase 1"
- Adjust with "add {{feature}}" or "postpone {{feature}}"