browser-navigation

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the agent-browser CLI to automate web interactions and browser management through shell commands.
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to install the agent-browser package globally or execute it via npx from the npm registry.
  • [DYNAMIC_EXECUTION]: The eval command provides a mechanism to execute arbitrary JavaScript code within the browser's execution environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by retrieving and analyzing content from external, untrusted websites.
  • Ingestion points: Untrusted data enters the agent context through page snapshots (snapshot), text extraction (get text), and console log monitoring (console).
  • Boundary markers: The skill does not implement specific delimiters or instructions to prevent the agent from interpreting content from web pages as instructions.
  • Capability inventory: The skill possesses capabilities for executing the agent-browser CLI, writing files (screenshots, PDFs, session state, and video recordings), and performing network operations via the browser.
  • Sanitization: There is no evidence of content sanitization or validation for data retrieved from web pages.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 04:38 PM
Security Audit — agent-trust-hub — browser-navigation