azure-ai-voicelive-py

Pass

Audited by Gen Agent Trust Hub on Aug 9, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a technical guide for using the azure-ai-voicelive SDK. The provided examples are consistent with standard development practices for voice-enabled AI applications.
  • [EXTERNAL_DOWNLOADS]: The installation section instructs the user to install azure-ai-voicelive, aiohttp, and azure-identity. These packages are necessary for the skill's functionality, and their usage aligns with the intended purpose of communicating with Azure services.
  • [CREDENTIALS_UNSAFE]: The skill documents the use of AZURE_COGNITIVE_SERVICES_KEY but explicitly notes that this method is "not recommended for production." It correctly prioritizes DefaultAzureCredential for more secure, identity-based authentication.
  • [DATA_EXFILTRATION]: Network operations are scoped to Microsoft's official cognitive services domain (api.cognitive.microsoft.com). There are no patterns suggesting data is being sent to unauthorized external servers.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes an ingestion surface for model-generated content via the WebSocket event loop and function calling (handle_function).
  • Ingestion points: Event stream from conn WebSocket.
  • Boundary markers: Explicit session instructions are set during conn.session.update.
  • Capability inventory: The skill can process audio transcripts and execute function calls defined in the tool schema.
  • Sanitization: Standard Base64 encoding is used for audio data transmission, though the provided code is a template and expects the developer to implement specific function handling logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 9, 2026, 06:29 PM
Security Audit — agent-trust-hub — azure-ai-voicelive-py