Aliyun Openclaw

提供阿里云服务器上OpenClaw网关的远程部署、SSH隧道连接、设备配对及日常管理全流程支持。

Security Scan

The skill's content is broadly consistent with a remote OpenClaw deployment guide, but it embeds many sensitive credentials, instructs risky actions (root SSH with plaintext password, copying local skill directories to a remote host, enabling open origins), and the manifest does not declare the required credentials — these mismatches and unsafe practices warrant caution.

Purpose & Capability

The name/description (deploy & manage OpenClaw on an Alibaba Cloud server) matches the instructions: SSH, Docker, starting services, and device pairing are expected. However the skill files contain hard-coded secrets and model/config credentials while the skill metadata declares no required env vars or credentials — a coherence issue and poor hygiene even if functionally plausible.

Instruction Scope

Runtime instructions ask the operator to: SSH as root using a plaintext password, use sshpass/scp to copy ~/.openclaw/workspace/skills (accessing local filesystem), run docker commands that mount/modify remote config and skills, and remove device files. These steps go beyond a narrow helper: they perform system-wide actions and give the skill (and synced skills) potential to run arbitrary code on the remote host.

Install Mechanism

This is instruction-only (no install spec) which is low platform-install risk. The guide asks the user to npm install/private package '@openclaw/browser-relay' inside the container; that package is noted as an internal (non-public) package — installing unverified/internal packages on a production host is a risk and may be hard to audit.

Credentials

Although the skill metadata lists no required env vars/credentials, the SKILL.md and included JSON files contain multiple sensitive secrets: an SSH root password, Gateway token, model API key (bailian), TAVILY API key, and Feishu appSecret/appId. Requesting or embedding multiple high-scope secrets (root password, API keys, app secrets) is disproportionate and not declared in the manifest.

Persistence & Privilege

The instructions tell the user to run the OpenClaw container with --restart always and to sync local skills into /app/skills on the remote — this grants persistent execution and the ability to install new skills (code) on the server. The remote gateway is configured with controlUi.allowedOrigins=["*"] and a static token in docs/URLs, which risks exposing the admin UI if network/firewall rules are not hardened.

Scan Findings in Context

[embedded_secret:ssh_password] unexpected: SKILL.md and browser_relay_guide.md include plaintext SSH credentials (root / Davinci@1984). A deployment guide might need SSH access, but embedding a root password in distributed skill files is insecure and the manifest did not declare this credential.

[embedded_secret:gateway_token] expected: A gateway token is required to configure and access the OpenClaw control UI; however it is hard-coded in multiple files and exposed in example URLs (http://localhost:18790/#token=...). Hard-coding tokens in shared skill files is unsafe.

[embedded_secret:api_key_bailian] expected: The bailian model API key appears in config JSON. Supplying a model API key is reasonable for integrated model access, but again it is embedded in distributed files instead of being declared as a required secret or stored in a secure secret store.

[embedded_secret:feishu_app_secret] expected: Feishu appSecret and appId are present in feishu config files. Channel integration needs these credentials, but embedding them in the skill bundle without manifest declaration is poor practice and increases exfiltration risk.

[embedded_secret:tavily_api_key] expected: TAVILY_API_KEY is present in multiple JSON files. It may be required for some integrations, but again should be declared and handled securely rather than hard-coded.

What to consider before installing

This skill appears to be a genuine deployment/config guide, but it contains multiple hard-coded sensitive credentials (root SSH password, gateway token, model API key, Feishu secrets) and instructs risky operations: using sshpass with a root password, copying your entire local skills directory to the remote host, and enabling control UI origins="*". Before using it: (1) do not copy or use the plaintext credentials — replace them with per-host SSH key auth and rotate any exposed keys/tokens; (2) inspect any local skills you plan to sync (they can run arbitrary code on the remote); (3) restrict gateway/control UI access with firewall rules and avoid allowedOrigins="*"; (4) verify the Docker image registry and any internal npm packages before installing; (5) store API keys and app secrets in a secure secrets manager rather than committing them into config files. Because the manifest does not declare required secrets yet the files include many, treat this skill with caution and consider remediation (remove secrets, use keys, limit copied content) before running any commands.

aliyun_openclaw Skill

阿里云 OpenClaw 远程部署与管理技能。提供远程服务器上的 OpenClaw 网关部署、SSH 隧道连接、设备配对和日常管理的完整流程。

触发词

• 阿里云 OpenClaw
• 远程 OpenClaw 部署
• OpenClaw 配对
• 飞书连接 OpenClaw
• SSH 隧道 OpenClaw

环境配置

服务器信息

项目	值
远程服务器 IP	47.115.54.84
SSH 用户名	root
SSH 密码	Davinci@1984
本地访问端口	18790
远程网关端口	18789
Gateway Token	f9df2ba3bd91e46e81d186b5f74457c7043085cb9a1df4a3

大模型配置 (已同步)

默认模型： bailian/qwen3.5-plus

可用模型列表：

模型 ID	名称
qwen3.5-plus	Qwen3.5 Plus (默认)
qwen3-max-2026-01-23	Qwen3 Max
qwen3-coder-next	Qwen3 Coder Next
qwen3-coder-plus	Qwen3 Coder Plus
MiniMax-M2.5	MiniMax M2.5
glm-5	GLM-5
glm-4.7	GLM-4.7
kimi-k2.5	Kimi K2.5

API 配置：

• Provider: bailian (阿里百炼)
• Base URL: https://coding.dashscope.aliyuncs.com/v1
• API Key: sk-sp-dc9e54ca53434d4a828241fb51b60f52

MCP 配置

注意： 远程服务器暂未安装 MCP 适配器插件。如需使用 TAPD 等 MCP 工具，需要：

1. 在远程服务器安装：docker exec openclaw npm install -g @openclaw/mcp-adapter

Browser Relay 配置

完整文档： skills/aliyun_openclaw/browser_relay_guide.md

快速配置：

# 1. 安装 Browser Relay（在阿里云执行）
ssh root@47.115.54.84 "docker exec openclaw sh -c 'cd /app && npm install @openclaw/browser-relay'"

# 2. 启动服务
ssh root@47.115.54.84 "docker exec -d openclaw sh -c 'cd /app && npx @openclaw/browser-relay --port 18792 &'"

# 3. 建立 SSH 隧道（本地执行）
ssh -f -N -L 18792:localhost:18792 root@47.115.54.84

# 4. 配置扩展
# Port: 18792
# Gateway token: f9df2ba3bd91e46e81d186b5f74457c7043085cb9a1df4a3

2. 或手动配置 MCP Server

已同步的 Skills

本地 Skills 目录：~/.openclaw/workspace/skills/

核心 Skills 列表：

• api-gateway - API 网关集成
• brave-search - 网络搜索
• browser-automation - 浏览器自动化
• frontend-design - 前端设计
• humanizer - AI 文本人性化
• mcp-adapter - MCP 适配器
• playwright-scraper-skill - 网页爬取
• stock-analysis - 股票分析
• tushare-data - Tushare 财经数据
• tavily-search - Tavily 搜索
• summarize - 内容摘要
• oa-check-in - OA 签到
• zentao - 禅道集成
• aliyun_openclaw - 本技能 (阿里云部署文档)

同步 Skills 到远程服务器：

# 打包本地 skills
cd ~/.openclaw/workspace/skills/
tar -czf skills.tar.gz */

# 传输到远程服务器
sshpass -p 'Davinci@1984' scp skills.tar.gz root@47.115.54.84:/opt/openclaw/skills/

# 在远程服务器解压
ssh root@47.115.54.84
cd /opt/openclaw/skills/
tar -xzf skills.tar.gz

目录结构

/opt/openclaw/
├── config/         # 配置文件 (openclaw.json)
├── workspace/      # 工作区
├── memory/         # 记忆文件
├── skills/         # 技能文件 (需手动同步)
└── logs/           # 日志文件

使用流程

一、首次部署（仅需执行一次）

# 1. SSH 登录远程服务器
ssh root@47.115.54.84

# 2. 创建配置目录
mkdir -p /opt/openclaw/{config,workspace,memory,skills,logs}

# 3. 上传配置文件 (从本地)
scp ~/.openclaw/workspace/skills/aliyun_openclaw/remote_config.json root@47.115.54.84:/opt/openclaw/config/openclaw.json

# 4. 启动 Docker 容器
docker run -d \
  --name openclaw \
  --restart always \
  -p 18789:18789 \
  -p 18791:18791 \
  -v /opt/openclaw/config:/home/node/.openclaw \
  -v /opt/openclaw/workspace:/app/workspace \
  -v /opt/openclaw/memory:/app/memory \
  -v /opt/openclaw/skills:/app/skills \
  -v /opt/openclaw/logs:/app/logs \
  registry.cn-hangzhou.aliyuncs.com/qiluo-images/openclaw:latest

# 5. 检查状态
docker logs openclaw --tail 10

二、本地连接（每次使用）

# 1. 建立 SSH 隧道（后台运行）
ssh -f -N -L 18790:localhost:18789 root@47.115.54.84

# 2. 验证隧道
curl -s -o /dev/null -w "%{http_code}" http://localhost:18790/
# 应返回 200

# 3. 访问仪表盘
# 浏览器打开：http://localhost:18790/#token=f9df2ba3bd91e46e81d186b5f74457c7043085cb9a1df4a3

三、设备配对流程

# 1. 查看待配对设备（在远程服务器执行）
docker exec openclaw openclaw devices list

# 2. 批准配对请求
docker exec openclaw openclaw devices approve <requestId>

# 3. 查看已配对设备
docker exec openclaw openclaw devices list

四、飞书集成

当用户在飞书 App 中连接 OpenClaw 时：

1. 飞书 App 会发起配对请求
2. 执行以下命令查看并批准：

# 查看待配对请求
docker exec openclaw openclaw devices list

# 批准飞书配对（替换 <requestId> 为实际 ID）
docker exec openclaw openclaw devices approve <requestId>

# 验证已配对
docker exec openclaw openclaw devices list

五、常用管理命令

# 查看网关状态
docker exec openclaw openclaw gateway status

# 查看日志
docker logs openclaw --tail 50

# 重启网关
docker restart openclaw

# 停止网关
docker stop openclaw

# 删除容器（重新配置时用）
docker stop openclaw && docker rm openclaw

# 更新到最新版本
docker pull registry.cn-hangzhou.aliyuncs.com/qiluo-images/openclaw:latest
docker restart openclaw

# 同步 Skills 到远程
cd ~/.openclaw/workspace/skills/
tar -czf skills.tar.gz */
sshpass -p 'Davinci@1984' scp skills.tar.gz root@47.115.54.84:/opt/openclaw/skills/
ssh root@47.115.54.84 "cd /opt/openclaw/skills/ && tar -xzf skills.tar.gz"

故障排查

SSH 隧道无法建立

# 检查是否有旧隧道进程
ps aux | grep "ssh.*18790"

# 杀掉旧进程
pkill -f "ssh.*18790"

# 重新建立隧道
ssh -f -N -L 18790:localhost:18789 root@47.115.54.84

网关无法访问

# 检查容器状态
docker ps | grep openclaw

# 查看容器日志
docker logs openclaw --tail 30

# 检查端口监听
docker exec openclaw ss -tlnp | grep 18789

配对失败

# 清除旧配对（谨慎使用）
docker exec openclaw sh -c 'rm -f /home/node/.openclaw/devices.json'
docker restart openclaw

模型配置问题

# 查看当前配置
docker exec openclaw cat /home/node/.openclaw/openclaw.json

# 修复配置
docker exec openclaw openclaw doctor --fix

安全提示

• ⚠️ SSH 密码、API Key 和 Gateway Token 属于敏感信息，不要公开分享
• ⚠️ 生产环境建议使用 SSH 密钥认证而非密码
• ⚠️ 建议配置防火墙限制 18789 端口仅允许信任 IP 访问
• ⚠️ 定期更新 OpenClaw 到最新版本
• ⚠️ 大模型 API Key 已同步到远程，注意用量监控

配置同步清单

配置项	本地	远程	状态
Gateway 配置	✅	✅	已同步
大模型配置	✅	✅	已同步 (bailian)
Skills	✅	⏳	需手动同步
MCP 配置	✅	❌	未安装插件
记忆文件	✅	⏳	需手动同步
工作区文件	✅	⏳	需手动同步

相关链接

• OpenClaw 文档：https://docs.openclaw.ai
• 设备配对文档：https://docs.openclaw.ai/web/control-ui#device-pairing-first-connection
• Docker 镜像：registry.cn-hangzhou.aliyuncs.com/qiluo-images/openclaw:latest
• 阿里百炼文档：https://help.aliyun.com/zh/model-studio/

配置文件备份

远程配置文件已保存在：skills/aliyun_openclaw/remote_config.json