browser-mcp-agent

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In this skill’s MCP runtime workflow, the agent uses tools like navigate plus get_content/evaluate to read page text from arbitrary URLs the operator supplies, and that page text is explicitly treated as untrusted input that the agent can ingest and use to choose subsequent tool calls.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's runtime config launches "npx anti-detect-browser@2.2.0 --mcp" which fetches and executes a remote npm package, and the browser kernel is downloaded at first launch from https://antibrow.com, so remote code/binary is fetched and executed during runtime.