本文へスキップ
TakoAPI
スキル一覧に戻る

speckit-swarm

heldinhow 作

Coding Agents & IDEs
441ダウンロード数はclawskills.shが公開している値です
0閲覧数
0件のいいねサインインするとこのスキルにいいねできます
—まだ評価がありません
あなたの評価

概要

Native implementation of oh-my-opencode-style orchestration using OpenClaw's tools.

説明

このスキルでできること

Agent orchestration system for OpenClaw that routes tasks to specialized AI personas and executes them in parallel or sequentially based on complexity and conflict analysis. Triggered by 'ulw'/'ultrawork' keywords for automatic parallel decomposition, or used manually by selecting a persona directly. Five built-in personas cover orchestration, deep implementation, design/debug, research, and fast scouting.

Automatically prevents concurrency conflicts by analyzing task type before parallelizing, unlike naive multi-agent setups that let agents overwrite each other.

使いどころ

  • Running parallel refactors across multiple unrelated modules
  • Assigning a bug hunt to the Oracle persona while Librarian looks up docs
  • Auto-decomposing a large feature build with 'ulw' prefix
  • Spawning a fast scout to locate files before deeper agents begin work
  • Sequencing dependent tasks to avoid concurrent file edit conflicts

ワークフロー例

入力

User asks: ulw create a new REST API with auth and tests

エージェント
  1. 1

    Ultrawork detector finds 'ulw' keyword and strips it from the task

  2. 2

    planTask() decomposes the request into parallel chunks: spec, setup, auth, tests

  3. 3

    Conflict analyzer checks each chunk for shared file dependencies

  4. 4

    parallel_spawn executes safe chunks concurrently using Sisyphus and Hephaestus personas

  5. 5

    Results are aggregated and returned to the user

出力

Parallel agents complete spec, scaffolding, auth logic, and test files independently, then results are combined

インストール

clawhub install heldinhow/speckit-swarm

CLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。

リンク