lobster-agent 安全扫描报告 — 低风险 | ClawSafe

10 /100

lobster-agent

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

This is a documentation-only skill package describing a legitimate server monitoring agent. The SKILL.md declares all major capabilities (network reporting, systemd service, root privileges) but contains no implementation code to verify actual behavior.

技能名称lobster-agent

分析耗时38.8s

引擎pi

✓

可以安装

Acceptable for use with low risk profile. However, since no implementation code exists in this package, verify that the actual lobster-agent code matches the documented behavior before deployment. Request source code review if available.

安全发现 2 项

严重性	安全发现	位置
低危	Implementation not included 文档欺骗 This skill package only contains SKILL.md documentation. No Python scripts or implementation files exist in the package. The actual lobster-agent code is not provided for security review. `All content is documentation only` → Request the actual implementation code before final security approval	`SKILL.md:1`
低危	External platform data reporting (declared) 数据外泄 The skill reports system metrics (CPU, memory, disk, network) to external Coze platform. This is the stated purpose of the skill. `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` → Verify data collection scope matches your organization's data policies	`SKILL.md:19`

资源类型	声明权限	推断权限	状态	证据
文件系统	`WRITE`	`WRITE`	✓ 一致	SKILL.md describes creating /opt/lobster-agent, /var/log/lobster-agent directori…
网络访问	`WRITE`	`WRITE`	✓ 一致	SKILL.md line 19: Reports to https://api.coze.cn
命令执行	`WRITE`	`WRITE`	✓ 一致	SKILL.md describes creating /usr/local/bin/lobster command and systemd service
环境变量	`NONE`	`NONE`	—	No environment access described
技能调用	`NONE`	`NONE`	—	No skill invocation described

3 项发现

🔗

中危外部 URL 外部 URL

https://api.coze.cn）

SKILL.md:19

🔗

中危外部 URL 外部 URL

https://www.coze.cn/docs/developer-docs/api

SKILL.md:137

🔗

中危外部 URL 外部 URL

https://coze.cn/s/7618478715609055278

SKILL.md:138

目录结构

1 文件 · 4.3 KB · 138 行

Markdown 1f · 138L

└─ 📝 SKILL.md Markdown 138L · 4.3 KB

安全亮点

✓ All major capabilities are documented in SKILL.md

✓ No obfuscation or base64-encoded content found

✓ No credential harvesting beyond monitoring requirements

✓ No reverse shell or C2 indicators

✓ External data transfer is declared as the core purpose

✓ Dependencies (requests, psutil) are reasonable for monitoring