agent-browser

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the agent-browser CLI and Playwright browser binaries. These resources originate from Vercel Labs, a recognized and trusted organization.- [COMMAND_EXECUTION]: The tool enables a wide range of browser interactions, including form filling, element clicks, and session management via persistent profiles.- [REMOTE_CODE_EXECUTION]: Includes the eval command, which allows the agent to execute arbitrary JavaScript code within the active browser environment.- [DATA_EXFILTRATION]: Provides capabilities to retrieve sensitive data such as cookies, local storage, and session storage. It also supports capturing page state through screenshots, PDFs, and video recordings.- [PROMPT_INJECTION]: The skill is inherently vulnerable to indirect prompt injection as it processes untrusted data from the web. \n
  • Ingestion points: agent-browser open <url> and agent-browser snapshot.\n
  • Boundary markers: None specified in the documentation.\n
  • Capability inventory: Significant, including browser-side script execution, network request interception, and file system interaction (uploads/downloads).\n
  • Sanitization: Not mentioned in the command set; the agent interprets the page DOM and accessibility tree directly.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 10:20 PM