gemini-live-api
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFE
Full Analysis
- Authentication Management: The skill correctly instructs the implementation to use Application Default Credentials (ADC) for bearer tokens rather than hardcoding sensitive API keys. It emphasizes token refresh and secure transport via WebSocket Authorization headers.
- Runtime Isolation: The skill follows best practices by requiring an isolated environment (e.g., Python venv) for generated code, explicitly forbidding mutations to the host's system environment or global packages.
- Secure Communication: All communication is directed to well-known Google Vertex AI/Gemini Enterprise endpoints (googleapis.com) over encrypted WebSockets (WSS). It also correctly identifiers appropriate project-scoped model paths.
- Remote Resource Verification: The skill references official Google Cloud documentation and GitHub repositories for its protocol definitions and best practices, aligning with trusted vendor resource patterns.
- Input Validation: The instructions include handling for real-time data modalities and explicit handling for interrupt signals, which are standard for low-latency bidirectional AI services.
Audit Metadata