agent-browser-harness
Pass
Audited by Gen Agent Trust Hub on Apr 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a collection of reference guides and instructions (markdown) designed to assist an agent in using the agent-browser CLI tool effectively. It does not contain executable scripts or code that performs suspicious actions.
- [SAFE]: The instructions encourage security best practices, such as stopping and requesting user intervention when encountering authentication walls, captchas, or other human-only checkpoints.
- [SAFE]: No hardcoded secrets, API keys, or credentials were found within the skill files.
- [SAFE]: Network activity and file uploads are handled through standard tool commands (e.g., agent-browser network requests, agent-browser upload) and are directed at legitimate targets described in the skill's purpose, such as TikTok Studio.
- [SAFE]: The skill documentation mentions the use of 'agent-browser eval' for state inspection, but explicitly categorizes it as a fallback method and warns against using it for DOM mutation, which reduces the risk associated with dynamic execution.
Audit Metadata