gemini-live-api
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFE
Full Analysis
- Implementation Guidelines: The skill provides structural instructions for generating code that connects to official Google Cloud Vertex AI endpoints. It correctly identifies the required authentication method (Application Default Credentials) and includes a checklist to ensure developers do not accidentally hardcode sensitive API keys.
- Environment Security: The instructions explicitly forbid mutating the host's system Python environment or using 'sudo pip install'. Instead, it mandates the creation of isolated virtual environments (venv) within the user's specified output directory, following security best practices for local development.
- External References: All external documentation links and WebSocket endpoints (e.g., aiplatform.googleapis.com) point to official Google Cloud and Vertex AI domains. These are trusted services for the intended use case of the skill.
- Authentication Management: The skill specifies that the generated client should use bearer tokens obtained via ADC and provides guidance on refreshing these tokens, which is a secure method for managing session-based authentication without exposing long-lived credentials.
Audit Metadata