agent-browser
Pass
Audited by Gen Agent Trust Hub on Jun 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues detected.
- [PROMPT_INJECTION]: The skill interacts with external web content, creating a surface for potential indirect prompt injection attacks. This risk is inherent to browser automation tools.
- Ingestion points: Web content is accessed via the 'agent-browser open' and 'snapshot' commands in 'SKILL.md'.
- Boundary markers: No explicit delimiters are defined to separate untrusted web content from system instructions.
- Capability inventory: The 'agent-browser' tool can perform navigation, click elements, fill forms, and save session state.
- Sanitization: No sanitization or validation of retrieved web content is described in the instructions.
Audit Metadata