parallel

PROLE-ISLAND's avatarfrom PROLE-ISLAND

並列タスク実行ヘルパー。複数タスクの同時実行・結果統合

0stars🔀0forks📁View on GitHub🕐Updated Jan 10, 2026

When & Why to Use This Skill

The Parallel Task Execution Helper is a high-performance Claude skill designed to optimize workflows by executing multiple independent tasks simultaneously. It features intelligent sub-agent routing based on task keywords (Frontend, Backend, DevOps, etc.) and supports advanced execution modes like 'all', 'race', and 'settle'. By integrating results into a comprehensive report, it significantly reduces total processing time and enhances productivity for complex, multi-step projects.

Use Cases

  • Full-Stack Development Acceleration: Simultaneously generate backend APIs, frontend components, and unit tests, reducing development time from minutes to seconds.
  • Competitive Research & Data Retrieval: Use 'race' mode to query multiple search engines or databases at once, adopting the fastest response for real-time information needs.
  • Automated System Auditing: Run parallel checks for security vulnerabilities, performance bottlenecks, and code style compliance across a repository to generate a unified health report.
  • Complex Workflow Orchestration: Automatically delegate specialized tasks to the most appropriate sub-agents (e.g., routing CSS tasks to a design agent and SQL tasks to a backend agent) for higher quality outputs.
nameparallel
description実行モード(all/race/settle)
- namemode
requiredfalse
default"all"

/parallel - 並列実行ヘルパー

複数の独立タスクを同時実行し、結果を統合する。

実行モード

all - 全タスク完了待ち

/parallel tasks="API実装,UI実装,テスト作成" mode=all
# → 全タスク完了後に結果統合

race - 最初の完了で終了

/parallel tasks="検索A,検索B,検索C" mode=race
# → 最初に完了したタスクの結果を採用

settle - 全タスク結果収集(失敗含む)

/parallel tasks="タスクA,タスクB,タスクC" mode=settle
# → 成功・失敗問わず全結果を収集

エージェント自動選択

タスク分析
  ↓
キーワード判定:
  - "UI", "コンポーネント", "フロント" → my-frontend
  - "API", "DB", "バックエンド" → my-backend
  - "Docker", "CI", "デプロイ" → my-devops
  - "デザイン", "スタイル", "UX" → my-design
  - その他 → general-purpose

実行フロー

// タスク分解
const tasks = parseTasks(input);

// エージェント割り当て
const assignments = tasks.map(task => ({
  task,
  agent: autoSelectAgent(task)
}));

// 並列実行
const results = await Promise.all(
  assignments.map(({ task, agent }) =>
    Task({
      subagent_type: agent,
      prompt: task,
      description: `Parallel: ${task.slice(0, 20)}...`
    })
  )
);

// 結果統合
return integrateResults(results);

結果レポート

## 並列実行レポート

### 実行サマリー
- 総タスク: 3
- 成功: 3
- 失敗: 0
- 実行時間: 45s(直列なら2m予想)

### タスク結果

| # | タスク | エージェント | 結果 | 時間 |
|---|-------|------------|------|------|
| 1 | API実装 | my-backend | ✅ | 20s |
| 2 | UI実装 | my-frontend | ✅ | 35s |
| 3 | テスト作成 | general | ✅ | 45s |

### 生成ファイル
- src/app/api/users/route.ts
- src/components/UserList.tsx
- src/lib/__tests__/users.test.ts

使用例

# 基本(自動エージェント選択)
/parallel tasks="ユーザーAPI,ユーザー一覧UI,ユーザーテスト"

# エージェント指定
/parallel tasks="機能A,機能B" agents="my-frontend,my-backend"

# レース実行
/parallel tasks="Google検索,Bing検索,DuckDuckGo検索" mode=race

# JSON形式
/parallel tasks='[{"name":"API","agent":"my-backend"},{"name":"UI","agent":"my-frontend"}]'
parallel – AI Agent Skills | Claude Skills