Scan Report
15 /100
browser-agent
浏览器自动化 Agent 技能,基于 Chrome DevTools Protocol (CDP) 和 WebSocket 实现 AI 对浏览器的直接控制
A legitimate browser automation skill using Chrome DevTools Protocol with all functionality documented. No malicious behavior detected; only local connections to localhost CDP endpoints.
Safe to install
Approve for use. Consider pinning dependency versions more strictly for production deployments.
Findings 2 items
| Severity | Finding | Location |
|---|---|---|
| Low | Dependency versions not strictly pinned Supply Chain | scripts/requirements.txt:2 |
| Low | package.json also uses flexible version ranges Supply Chain | package.json:33 |
| Resource | Declared | Inferred | Status | Evidence |
|---|---|---|---|---|
| Browser | READ | READ | ✓ Aligned | CDP WebSocket connections for browser control |
| Filesystem | WRITE | WRITE | ✓ Aligned | browser_agent.py:121 - writes screenshots to disk |
| Network | READ | READ | ✓ Aligned | browser_agent.py:64-70 - localhost HTTP/WebSocket only |
| Shell | NONE | NONE | — | No shell execution observed |
| Environment | NONE | NONE | — | No environment variable access detected |
4 findings
Medium External URL 外部 URL
https://chromedevtools.github.io/devtools-protocol/ README.md:190 Medium External URL 外部 URL
https://websocket-client.readthedocs.io/ README.md:191 Medium External URL 外部 URL
https://www.iana.org/help/example-domains 测试报告.md:138 Medium External URL 外部 URL
http://127.0.0.1:18800 测试报告.md:170 File Tree
9 files · 38.7 KB · 1425 lines Markdown 5f · 869L
Python 2f · 491L
JSON 1f · 55L
Text 1f · 10L
├─
▾
references
│ └─
api_reference.md
Markdown
├─
▾
scripts
│ ├─
browser_agent.py
Python
│ ├─
requirements.txt
Text
│ └─
session_manager.py
Python
├─
package.json
JSON
├─
README.md
Markdown
├─
SKILL.md
Markdown
├─
封装总结.md
Markdown
└─
测试报告.md
Markdown
Dependencies 2 items
| Package | Version | Source | Known Vulns | Notes |
|---|---|---|---|---|
websocket-client | >=1.6.0 | pip | No | Version not pinned |
requests | >=2.31.0 | pip | No | Version not pinned |
Security Positives
✓ All functionality explicitly declared in SKILL.md
✓ No credential harvesting or exfiltration detected
✓ Only local network connections (localhost:18800/9222) - no external IP communication
✓ No obfuscation, base64 execution, or suspicious encoding patterns
✓ No hidden functionality beyond documented features
✓ JavaScript evaluation (Runtime.evaluate) is standard browser automation, properly documented
✓ No access to sensitive paths (~/.ssh, ~/.aws, .env)
✓ No reverse shell, C2, or data theft patterns observed
✓ Clean code with no anti-analysis techniques