browser-use-cloud

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses Playwright and the Browser Use SDK to execute browser commands on a remote cloud instance. These commands (navigation, screenshots, recording) are core functionalities and are initiated by the agent or user.
  • [EXTERNAL_DOWNLOADS]: The skill downloads recordings (MP4) from the service provider's S3-backed URLs (api.browser-use.com) and installs standard dependencies (browser-use-sdk, playwright). These are expected behaviors for the described use case.
  • [DATA_EXFILTRATION]: Network communication is restricted to the official service domain (browser-use.com) for API interactions and remote browser control. No unauthorized transmission of sensitive user data was found.
  • [PROMPT_INJECTION]: The instructions describe the registration flow and usage of the tools without attempting to override system prompts or bypass safety filters.
  • [INDIRECT_PROMPT_INJECTION]: As a browser-based skill, it processes untrusted content from the web. However, it uses industry-standard automation tools (Playwright) which operate within the browser sandbox, and the skill does not contain patterns that would lead to host-level compromise via processed data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 07:09 AM
Security Audit — agent-trust-hub — browser-use-cloud