360 Secure SkillHub
360 安全技能中心

Calling Agent Squad

通过“calling squad [项目] [任务]”触发多代理小队(经理、架构师、研究员、文案、代码审查、品牌审查、观察员)。标准模式由当前代理依次代入各角色人设执行,完整模式经 openclaw 命令生成独立子代理,交付物统一保存到项目文件夹。

arbiger
arbiger
热度 56.44
最近更新 2026-05-18 12:26:48
分类:协作依赖:读写文件 / 代码执行 / 内部调用大模型应用场景:工作流 / 任务管理 / 研究 / 代码审查 ...

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add calling-agent-squad
Metadata
name
calling-agent-squad
version
1.0.0
description
Activate a multi-agent team (the Squad) to manage complex projects, business tasks, or development workflows. The squad includes a Manager, Architect, Coder, Reviewer, and Observer. Use when the user wants to "call a squad", "start a project", or "deploy squad" with specialized roles and quality control loops.
tags
["multi-agent","squad","workflow","project-management"]
author
Megan

Calling Agent Squad

This skill coordinates a specialized team of agents to handle your tasks with professional rigor.

🎯 Usage

Mode 1: Standard (Default)

calling squad [project] [task details]
  • I act as all roles (Manager, Researcher, Architect, Copywriter, Reviewer, Observer)
  • Before each role: I read their SOUL.md and IDENTITY.md to adopt their persona
  • After task completion: I return to my normal self (Megan)
  • Like playing a script - put on the mask, do the job, take it off

Mode 2: Full (Spawn Real Sub-Agents)

calling squad full [project] [task details]
  • Spawns real sub-agents via openclaw agent
  • Each agent runs in its own workspace with its own SOUL.md/IDENTITY.md
  • Slower but more professional - agents work independently
  • Suitable for complex tasks requiring specialized expertise

How It Works

Standard Mode

When user says calling squad [project] [task]:

  1. Create project folder: Documents/squad_projects/[project]_[yyyymmdd]/
  2. For each role (Researcher → Architect → Copywriter → Reviewer → Observer):
    • Read that role's SOUL.md and IDENTITY.md from agents/[role]/
    • Adopt their persona and complete their task
  3. Save all deliverables to project folder
  4. Return to normal (Megan) after completion

Full Mode

When user says calling squad full [project] [task]:

  1. Run: openclaw agent --agent squad-manager -m "Mission: [project] - [task]"
  2. Squad-manager spawns sub-agents with their own workspaces
  3. Each agent reads its own SOUL.md/IDENTITY.md
  4. Results saved to project folder

⚠️ Important Notes

  • Standard mode: I read SOUL/IDENTITY for each role, then return to Megan after - no memory contamination
  • Full mode: Sub-agents have separate context windows, completely isolated
  • Cost: Standard uses ~same tokens as normal; Full uses more (multiple agent sessions)

🛠️ Maintenance

To re-initialize agents (for Full mode):

bash ~/.openclaw/workspace/skills/calling-agent-squad/squad-init.sh

Folder Structure

Root: ~/.openclaw/workspace/skills/calling-agent-squad/agents/

Agent Config Folder
🦞 Manager agents/squad-manager/
📐 Architect agents/architect/
🔍 Researcher agents/researcher/
✍️ Copywriter agents/copywriter/
🛠️ Coder agents/coder/
🛡️ Reviewers agents/code-reviewer/, agents/brand-reviewer/
📋 Observer agents/observer/

Each folder contains: SOUL.md, IDENTITY.md, TOOLS.md, USER.md


Project Output

All projects saved to: Documents/squad_projects/[project]_[yyyymmdd]/


The Team

  • Squad Manager: Orchestrates, delegates, and arbitrates
  • Architect: Plans system blueprints and maintains handbook
  • Researcher: Gathers market and technical intelligence (facts first, deep insights)
  • Copywriter: Creates marketing and technical copy
  • Code Reviewer: Audits for security and logic errors
  • Brand Reviewer: Ensures brand consistency
  • Observer: Logs mission and extracts new rules