技能详情（站内镜像，无评论）
许可证：MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本：v1.0.0
统计：⭐ 0 · 38 · 0 current installs · 0 all-time installs
⭐ 0
安装量（当前） 0
🛡 VirusTotal ：良性 · OpenClaw ：可疑
Package：arbiger/calling-agent-squad
安全扫描（ClawHub）
- VirusTotal ：良性
- OpenClaw ：可疑
OpenClaw 评估
The skill's files and instructions broadly match a multi‑agent project orchestration purpose, but there are inconsistent/overbroad runtime directives (automatic file reads, commits/pushes, and a reusable init script) that warrant caution before installing or running.
目的
Name/description align with the provided files: this is a multi‑agent orchestrator with role personas and workspace conventions. It does not request external credentials or unusual binaries, which is proportionate. However the agents' manuals repeatedly instruct agents to read local memory files, edit MEMORY.md, and perform git commit/push operations — capabilities that are plausible for a squad orchestrator but are higher‑impact than the simp…
说明范围
SKILL.md and many AGENTS.md files instruct agents to read SOUL.md/IDENTITY.md and various memory files (memory/YYYY‑MM‑DD.md, MEMORY.md), create project folders under Documents/squad_projects, and the maintenance docs point to running bash ~/.openclaw/.../squad-init.sh. Some agent docs say 'Don't ask permission. Just do it.' and also permit 'commit and push your own changes' and running exec/tests. Those directives allow nontrivial local file …
安装机制
No install spec — instruction-only skill plus bundled files. That limits remote code downloads. There is a local initialization script (squad-init.sh) included; it will only run if the user explicitly executes it (e.g., maintenance instruction). Inspect the script before running.
证书
The skill declares no required env vars or credentials (proportionate). Nevertheless, many TOOLS.md examples mention SSH hosts, TTS, and other local secrets as optional notes; the skill's documents encourage storing local infra details in TOOLS.md. Those are examples rather than requirements, but users should avoid putting secrets into shared skill files.
持久
always:false and no install hooks are good. The skill expects and documents spawning sub‑agents (openclaw agent calls) and using platform operations like sessions_spawn/exec. Autonomous invocation is the platform default — not alone a concern — but combined with the agent directives to modify files, run commands, and push commits, it increases blast radius if misused. The included init script can be run manually; it does not appear to force pe…
安装（复制给龙虾 AI）
将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「Calling Agent Squad」。简介：Activate a multi-agent team (the Squad) to manage complex projects, business ta…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/arbiger/calling-agent-squad/SKILL.md
（来源：yingzhi8.cn 技能库）SKILL.md
---
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
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