azure-ai-voicelive-java
Pass
Audited by Gen Agent Trust Hub on Feb 14, 2026
Risk Level: LOW
Full Analysis
- [Trusted Source] (INFO): The skill references Maven dependencies (com.azure:azure-ai-voicelive) and the Azure GitHub organization, which are verified trusted sources. This finding is downgraded per [TRUST-SCOPE-RULE].
- [Indirect Prompt Injection] (LOW): The skill creates an interaction surface for untrusted voice data and supports function calling. (1) Ingestion points: Audio enters via 'session.sendInputAudio' in SKILL.md. (2) Boundary markers: No delimiters are used for the binary audio stream. (3) Capability inventory: Provides real-time audio interaction and tool calling ('VoiceLiveFunctionDefinition'). (4) Sanitization: Relies on the underlying Azure AI service for content filtering and safety.
- [Credentials Unsafe] (SAFE): The implementation correctly demonstrates the use of environment variables and 'DefaultAzureCredential', explicitly advising against hardcoded secrets.
Audit Metadata