Skills
skills/linuxhsj/openclaw-zero-token/browser-cdp/Gen Agent Trust Hub

browser-cdp

Fail

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: HIGHPROMPT_INJECTIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/cdp_proxy.py uses subprocess to locate and launch browser instances. These calls use argument lists, which is a safe practice that prevents shell injection attacks.
  • [REMOTE_CODE_EXECUTION]: The SKILL.md file contains instructions for piping JSON output to python3 -m json.tool. While automated scanners may flag this pattern, it is a benign use of a standard Python module for formatting JSON output.
  • [REMOTE_CODE_EXECUTION]: The /eval endpoint in scripts/cdp_proxy.py allows for direct JavaScript execution in the browser context. This is the intended functionality of the skill, though the local proxy does not implement access control for its API.
  • [PROMPT_INJECTION]: The skill facilitates indirect prompt injection by design as it ingests untrusted data from the internet.
  • Ingestion points: Page content and metadata accessed via scripts/cdp_proxy.py.
  • Boundary markers: None.
  • Capability inventory: Browser navigation, JavaScript execution, and screenshot capture in scripts/cdp_proxy.py.
  • Sanitization: No sanitization of the retrieved web content is performed.
Recommendations
  • HIGH: Downloads and executes remote code from: http://localhost:3456/targets - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 8, 2026, 05:11 PM