# Cross Agent

> 跨机器Agent协作 - 通过SSH连接局域网内其他OpenClaw实例，实现多机任务分发

- Skill: `dvcrn/cross-agent` (Agent Skill, multi-file: 13 files)
- Install (CLI): `npx skillmds@latest add dvcrn/cross-agent`
- Raw SKILL.md: https://api.skillmd.com/api/skills/dvcrn/cross-agent/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: dvcrn (https://skillmd.com/u/dvcrn)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/dvcrn/cross-agent

---


# 🔌 Cross-Agent 跨机器Agent协作

通过SSH连接到局域网内其他运行OpenClaw的机器，向其Agent发送任务指令。

## 🚀 快速开始

### 1. 扫描局域网设备
```bash
openclaw cross-agent scan 192.168.3.0/24
```

### 2. 测试SSH连接
```bash
openclaw cross-agent test 192.168.3.54 admin 123456
```

### 3. 发送任务
```bash
openclaw cross-agent send 192.168.3.54 "请搜索K230部署方案" admin 123456
```

### 4. 获取结果文件
```bash
openclaw cross-agent get 192.168.3.54 ~/Desktop/K230_部署方案.md ~/Desktop/ admin 123456
```

## ⚙️ 配置默认值

设置常用参数，后续命令可省略：
```bash
openclaw cross-agent config \
  --default-user admin \
  --default-pass 123456 \
  --default-ip 192.168.3.54
```

配置后简写：
```bash
openclaw cross-agent send "请搜索Docker教程"
```

## 📋 前提条件

### 目标机器必须：
1. ✅ 开启SSH服务（端口22）
2. ✅ 运行OpenClaw Gateway
3. ✅ 与本机在同一局域网

### 本机必须：
1. ✅ 安装sshpass: `sudo apt install sshpass`
2. ✅ 网络连通: `ping <目标IP>`

## 🔧 完整工作流示例

```bash
# 1. 扫描找到目标机器
openclaw cross-agent scan

# 2. 配置默认参数
openclaw cross-agent config --default-ip 192.168.3.54

# 3. 测试连接
openclaw cross-agent test

# 4. 查看目标会话
openclaw cross-agent sessions

# 5. 发送任务
openclaw cross-agent send "请生成一份Python爬虫教程"

# 6. 等待几秒...
sleep 10

# 7. 获取生成的文件
openclaw cross-agent get ~/Desktop/*.md ~/Desktop/

# 8. 查看内容
cat ~/Desktop/*.md
```

## 🐛 故障排除

| 问题 | 解决 |
|------|------|
| Permission denied | 检查用户名/密码 |
| Connection refused | 目标SSH未开启或防火墙阻挡 |
| Gateway token error | 使用scp/get/put传输文件 |
| Command not found | 目标机器未安装OpenClaw |

## 📚 更多帮助

- `openclaw cross-agent --help` - 查看所有命令
- `openclaw cross-agent <command> --help` - 查看具体命令帮助