skills/openai/plugins/vercel-sandbox/Gen Agent Trust Hub

vercel-sandbox

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFE
Full Analysis
  • Ephemeral Sandboxing: The skill documents the use of Vercel Sandbox (Firecracker microVMs) to execute code in isolation. This is a standard security practice for running untrusted code while protecting the host environment.
  • Authentication Management: The implementation examples demonstrate proper handling of credentials by checking for environment variables (VERCEL_TOKEN, VERCEL_TEAM_ID, VERCEL_PROJECT_ID) and utilizing Vercel's automatic OIDC authentication in production environments.
  • System Dependency Management: The skill provides a clear list of required system libraries for Chromium and uses standard package managers (dnf, npm) to install them within the isolated VM.
  • Snapshot Functionality: It includes instructions for creating and using sandbox snapshots, which is an optimization for fast startup times without introducing new security risks.
  • Browser Automation Controls: The patterns for browser automation include explicit commands to open, interact with, and close the browser instance, ensuring resources are cleaned up after execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 05:55 AM
Security Audit — agent-trust-hub — vercel-sandbox