agent-browser
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Installs the agent-browser CLI tool from the global NPM registry if not already present.
- [COMMAND_EXECUTION]: Executes shell commands to drive browser sessions, including page navigation, element interaction, and diagnostic capture. It also executes a capture helper via the $CAPTURE environment variable.
- [PROMPT_INJECTION]: The skill processes untrusted content from external web pages, exposing the agent to potential indirect prompt injection attacks where malicious web content could influence the agent's behavior.
- Ingestion points: Browser accessibility tree data, console logs, and network request data retrieved from arbitrary URLs.
- Boundary markers: Absent. The skill does not instruct the agent to ignore or delimit instructions found within the browser's output.
- Capability inventory: The skill has capabilities for network access, file system writes (screenshots), and CLI-based command execution.
- Sanitization: None; data retrieved from web pages is directly processed in the journey loop.
- [REMOTE_CODE_EXECUTION]: The command 'agent-browser skills get core' is used to dynamically fetch command references and guides from a remote source at runtime.
Audit Metadata