← スキル一覧に戻る

plan-swarm
by Uniswap
🤖 AI Toolkit - Standardized setup for Claude Code AI workflows. Nx monorepo with reusable agents, commands, and generators for enhanced AI-assisted development.
⭐ 13🍴 2📅 2026年1月24日
SKILL.md
description: Refine plans through multi-agent expert discussion. Use when user says "get multiple expert opinions on this plan", "I want different perspectives on this approach", "have the agents discuss this strategy", "refine the plan through collaborative review", or "what would different specialists think about this design". allowed-tools: Read, Glob, Grep, Task, WebSearch, WebFetch model: opus
Plan Swarm
Refine plans through collaborative multi-agent expert discussion.
⚠️ Cost Warning: This skill uses the Opus model and spawns 3-10 agents for multi-round discussions. A typical plan-swarm session consumes 5-20x more tokens than a standard skill invocation. Consider using
review-planfor simpler validation needs.
When to Activate
- User wants multiple expert perspectives
- Plan needs collaborative refinement
- Complex trade-offs require discussion
- Consensus building needed
- User asks for "expert opinions" or "different perspectives"
Key Features
- Intelligent Agent Selection: 3-10 agents based on plan context
- True Collaboration: Multi-round discussions
- Constructive Disagreement: Respectful challenging of ideas
- Consensus Building: Resolution of conflicts
Quick Process
Phase 1: Context & Agent Selection
- Analyze plan domains (frontend, backend, security, etc.)
- Select 3-10 relevant specialized agents
- Brief each agent on the plan
Phase 2: Multi-Round Discussion
- Round 1: Initial perspectives from each agent
- Round 2: Cross-domain discussion and response
- Round 3: Consensus building (if needed)
Phase 3: Synthesis
- Compile feedback across rounds
- Document consensus recommendations
- Identify trade-offs and open questions
Agent Selection Guidelines
| Plan Complexity | Agents |
|---|---|
| Simple | 3-4 |
| Medium | 5-7 |
| Complex | 8-10 |
Output Format
Generates refinement document with:
- Executive summary
- Agent participants and focus areas
- Consensus recommendations by category
- Design decisions and trade-offs
- Open questions requiring human decision
- Dissenting opinions with rationale
- Next steps
Discussion Guidelines
Agents are encouraged to:
- Be direct and state opinions clearly
- Challenge constructively with alternatives
- Build on other agents' ideas
- Change positions when persuaded
- Acknowledge expertise limits
スコア
総合スコア
70/100
リポジトリの品質指標に基づく評価
✓SKILL.md
SKILL.mdファイルが含まれている
+20
✓LICENSE
ライセンスが設定されている
+10
✓説明文
100文字以上の説明がある
+10
○人気
GitHub Stars 100以上
0/15
○最近の活動
3ヶ月以内に更新がある
0/10
○フォーク
10回以上フォークされている
0/5
✓Issue管理
オープンIssueが50未満
+5
✓言語
プログラミング言語が設定されている
+5
○タグ
1つ以上のタグが設定されている
0/5
レビュー
💬
レビュー機能は近日公開予定です