← 返回 Skills 市场
259
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install lobster-agent
功能描述
服务器监控Agent,自动采集系统指标并上报到Coze大龙虾平台,支持CPU/内存/磁盘/网络监控、告警推送和自动节点注册。
使用说明 (SKILL.md)
🦞 小龙虾监控Agent Skill (lobster-agent)
Overview
此Skill为服务器监控代理,可自动采集系统运行指标并上报到Coze大龙虾平台,具备以下核心功能:
- 指标采集 – CPU使用率、负载、内存使用率、磁盘使用率、网络流量等。
- 自动告警 – 当指标超过阈值时自动产生告警(支持warning/critical两级)。
- 节点注册 – 安装时自动注册到Coze平台,生成唯一Node ID。
- 数据上报 – 心跳包、监控数据、告警数据分别上报到对应数据集。
- 系统服务 – 后台运行,开机自启,崩溃自动重启。
Prerequisites
- 操作系统:Linux(支持systemd)
- 依赖:Python3、pip3、curl(系统默认一般已安装)
- 网络:可访问Coze API(https://api.coze.cn)
- 权限:root权限(安装系统服务需要)
Usage Steps
1. 安装Agent
直接运行内置安装脚本即可完成全自动化安装:
# 自动执行安装流程(OpenClaw触发)
安装过程会自动完成:
- 环境检查和依赖安装(requests、psutil)
- 工作目录创建(/opt/lobster-agent、/var/log/lobster-agent)
- Agent主程序生成
- 节点自动注册,获取Node ID
- 配置文件生成
- 管理命令创建(/usr/local/bin/lobster)
- systemd服务创建并启动
2. 管理Agent
安装完成后可通过lobster命令管理服务:
# 查看服务状态
lobster status
# 查看实时日志
lobster logs
# 启动服务
lobster start
# 停止服务
lobster stop
# 重启服务
lobster restart
# 完全卸载
lobster uninstall
3. 配置说明
配置文件路径:/opt/lobster-agent/config.json
{
"node_id": "自动生成的节点ID",
"api_key": "Coze API密钥",
"coze_base_url": "Coze API地址",
"dataset_ids": {
"nodes": "节点信息数据集ID",
"monitor_data": "监控数据数据集ID",
"alerts": "告警数据数据集ID"
},
"heartbeat_interval": 300, // 心跳间隔(秒)
"report_interval": 1800, // 监控数据上报间隔(秒)
"enable_hardware_check": true,
"enable_log_check": true
}
4. 告警阈值
| 指标 | Warning阈值 | Critical阈值 |
|---|---|---|
| CPU使用率 | >80% | >90% |
| 内存使用率 | >85% | >95% |
| 磁盘使用率 | >70% | >90% |
Example
安装成功输出
🎉 ==============================================
✅ 小龙虾Agent安装成功!
📊 Node ID: 7618478715609055278
📝 管理命令: lobster [status|start|stop|restart|logs|uninstall]
📜 查看日志: lobster logs
🌐 数据已自动上报到Coze平台
🎉 ==============================================
查看状态
lobster status
输出示例:
● lobster-agent.service - Lobster Monitor Agent
Loaded: loaded (/etc/systemd/system/lobster-agent.service; enabled; vendor preset: enabled)
Active: active (running) since Thu 2026-03-19 08:41:23 CST; 10min ago
Main PID: 12345 (python3)
Tasks: 1 (limit: 4915)
Memory: 20.0M
CGroup: /system.slice/lobster-agent.service
└─12345 /usr/bin/python3 /opt/lobster-agent/main.py
Data Structure
上报的监控数据字段
| 字段 | 类型 | 说明 |
|---|---|---|
| node_id | string | 节点唯一标识 |
| timestamp | string | 上报时间(YYYY-MM-DD HH:MM:SS) |
| cpu_usage | float | CPU使用率(%) |
| memory_usage | float | 内存使用率(%) |
| disk_usage | float | 磁盘使用率(%) |
| load_1 | float | 1分钟负载 |
| load_5 | float | 5分钟负载 |
| load_15 | float | 15分钟负载 |
| network_in | float | 入站流量(KB/s) |
| network_out | float | 出站流量(KB/s) |
| status | string | 节点状态(healthy/warning/critical) |
| alerts | string | 告警列表(JSON格式) |
When Not to Use
- 非Linux操作系统(不支持systemd)
- 没有root权限的环境
- 无法访问Coze API的离线环境
- 仅需要本地监控不需要云端上报的场景
References
- Coze API文档:https://www.coze.cn/docs/developer-docs/api
- 大龙虾监控平台:https://coze.cn/s/7618478715609055278
安全使用建议
Do not install this on production or sensitive hosts until you verify provenance. Key questions to ask the publisher: Where is the agent source code or release artifact (GitHub/official domain)? Can you provide checksums/signatures for binaries? Provide an install spec that explicitly downloads a named release from a trusted URL, or supply the full install script for audit. Confirm exactly what host data is sent to https://api.coze.cn and how API keys are stored/encrypted. If you need to test, run inside an isolated VM or container with no sensitive data and restricted network access. If you cannot get a trustworthy source and verifiable install artifacts, treat this skill as risky.
功能分析
Type: OpenClaw Skill
Name: lobster-agent
Version: 1.0.0
The skill bundle describes a server monitoring agent but lacks the actual implementation code, instead providing instructions in SKILL.md for the AI agent to 'automatically execute' a high-privilege installation process. This process requires root access, creates systemd services for persistence, and reports system metrics to an external endpoint (api.coze.cn). The absence of the source code for the installation script and the main agent program, combined with instructions for the AI to generate and execute system-level changes, presents a significant security risk without clear evidence of malicious intent.
能力评估
Purpose & Capability
The SKILL.md describes installing a systemd-backed monitoring agent (requires root, Python3, pip3, curl, network access to https://api.coze.cn). The registry metadata lists no required binaries, no env vars, and no install source. That mismatch suggests the declared capabilities/requirements do not match what the skill will actually do.
Instruction Scope
Runtime instructions say the agent will create /opt/lobster-agent, /var/log/lobster-agent, /usr/local/bin/lobster, and an /etc/systemd/system service, auto-register the node and send metrics/alerts to Coze. The SKILL.md provides no concrete, auditable install script or source for the agent binary; it authorizes collection and external transmission of host metrics, which is sensitive. Instructions do not reference unrelated files, but they do permit broad actions on the host.
Install Mechanism
There is no install spec and no code files — yet the doc claims the agent program will be generated/installed. Because there is no declared source (no URL, repo, or packaged artifact), it's unclear where executable code would come from. That absence increases risk: installation might require downloading unspecified code or generating it dynamically.
Credentials
The skill will store an api_key and other config under /opt/lobster-agent/config.json and requires root to install, but the registry metadata lists no required credentials or config paths. Requesting root and sending host metrics to an external service is a privileged operation; the lack of declared env/config requirements is incoherent and hides needed secrets (Coze API key).
Persistence & Privilege
The agent is designed to install a systemd service that runs on boot and restarts on crash (persistent, privileged presence). The skill metadata does not declare this persistence. The skill is user-invocable and can be invoked autonomously by the model (platform default); combined with service persistence and network exfiltration, that increases potential impact if the code or remote endpoint are untrusted.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install lobster-agent - 安装完成后,直接呼叫该 Skill 的名称或使用
/lobster-agent触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
lobster-agent 1.0.0 – Initial Release
- 首次发布,支持采集服务器CPU、内存、磁盘、网络等关键指标,并自动上报到Coze大龙虾平台。
- 内置自动告警机制,指标超过阈值时,分warning/critical两级推送告警。
- 安装时自动注册节点,生成唯一Node ID,支持全自动部署与systemd服务管理。
- 提供lobster命令工具,方便管理Agent服务与日志。
- 数据结构和配置文件明确,支持自定义阈值及多项监控设置。
元数据
常见问题
Lobster Agent 是什么?
服务器监控Agent,自动采集系统指标并上报到Coze大龙虾平台,支持CPU/内存/磁盘/网络监控、告警推送和自动节点注册。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 259 次。
如何安装 Lobster Agent?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install lobster-agent」即可一键安装,无需额外配置。
Lobster Agent 是免费的吗?
是的,Lobster Agent 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Lobster Agent 支持哪些平台?
Lobster Agent 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 Lobster Agent?
由 hdguodada(@hdguodada)开发并维护,当前版本 v1.0.0。
推荐 Skills