agent-browser

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute the agent-browser CLI on the host system. This tool is used to interact with web pages, capture screenshots, and record videos of browser sessions. The execution occurs outside the standard sandbox environment to allow interaction with local services.
  • [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by processing external web content. When the agent uses the snapshot -i command, it ingests the DOM structure and text of the target web page, which may contain malicious instructions designed to manipulate the agent's logic or subsequent interactions.
  • Ingestion points: External content enters the agent's context through the snapshot -i command in SKILL.md.
  • Boundary markers: The instructions do not define specific delimiters or warnings to help the agent distinguish between page content and its own instructions.
  • Capability inventory: The agent can perform interactive actions (click, fill, press), record video, and take screenshots based on the ingested content.
  • Sanitization: There are no explicit sanitization or filtering steps mentioned for the data retrieved from the browser snapshot.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 01:20 AM
Security Audit — agent-trust-hub — agent-browser