skills/jybill/xqv-skills/next-browser/Gen Agent Trust Hub

next-browser

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill installs the @vercel/next-browser package and Chromium using Playwright. These are sourced from well-known and trusted technology organizations.
  • [COMMAND_EXECUTION]: The skill uses a set of CLI commands to interact with the browser, inspect React component trees, and measure performance metrics. These commands are localized to the development environment.
  • [PROMPT_INJECTION]: The skill explicitly identifies the risk of indirect prompt injection from web page content (e.g., DOM attributes, console logs). It instructs the agent to treat page content as data and never as instructions, providing a mitigation layer for this attack surface.
  • [DATA_EXFILTRATION]: The skill manages authentication tokens and session cookies required for debugging authenticated pages. It implements a 'file-path-only' policy where secrets are stored in local files provided by the user, and the agent is forbidden from reading, echo-ing, or displaying the secret values directly.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 01:46 AM
Security Audit — agent-trust-hub — next-browser