扫描报告
5 /100
open-browser
Visual AI browser automation via OpenBrowser Agent — sees pages via screenshots and simulates human browser interactions
OpenBrowser is a legitimate visual AI browser automation skill with no malicious behavior — all operations are local, subprocess usage is documented, and no credentials are exfiltrated.
可以安装
No action needed. The skill is safe to use as described.
| 资源类型 | 声明权限 | 推断权限 | 状态 | 证据 |
|---|---|---|---|---|
| 文件系统 | READ | READ | ✓ 一致 | Scripts read from local server only; no arbitrary file writes without user conse… |
| 网络访问 | READ | READ | ✓ 一致 | All HTTP/WebSocket calls target localhost:8765; DashScope API is the intended AI… |
| 命令执行 | WRITE | WRITE | ✓ 一致 | subprocess.Popen used for background task execution, explicitly documented in SK… |
| 环境变量 | READ | READ | ✓ 一致 | OPENBROWSER_CHROME_UUID read from env, used only for local server registration |
| 浏览器 | WRITE | WRITE | ✓ 一致 | Chrome extension browser control via WebSocket, the core purpose of the skill |
4 项发现
中危 外部 URL 外部 URL
https://dashscope.aliyun.com/ SKILL.md:111 中危 外部 URL 外部 URL
http://127.0.0.1:8765 references/api_reference.md:7 中危 外部 URL 外部 URL
https://dashscope.aliyuncs.com/compatible-mode/v1 references/api_reference.md:56 中危 外部 URL 外部 URL
https://docs.astral.sh/uv/ references/setup.md:7 目录结构
6 文件 · 34.5 KB · 1280 行 Markdown 4f · 778L
Python 2f · 502L
├─
▾
references
│ ├─
api_reference.md
Markdown
│ ├─
setup.md
Markdown
│ └─
troubleshooting.md
Markdown
├─
▾
scripts
│ ├─
check_status.py
Python
│ └─
send_task.py
Python
└─
SKILL.md
Markdown
安全亮点
✓ All network operations are local (127.0.0.1:8765) except DashScope API for browser automation
✓ No credential harvesting or environment variable iteration for sensitive keys
✓ No base64, eval, or obfuscated code patterns
✓ No access to ~/.ssh, ~/.aws, .env, or other sensitive host paths
✓ No data exfiltration or C2 communication
✓ Subprocess usage is explicitly declared and documented in SKILL.md
✓ Browser UUID is a local capability token — not a secret, scoped to the local server
✓ No supply chain risks — all dependencies are from known package managers with pinned tooling
✓ Scripts are readable, well-structured Python with no hidden functionality