azure-ai-voicelive-py

Pass

Audited by Gen Agent Trust Hub on Mar 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the azure-ai-voicelive package along with aiohttp and azure-identity via pip. While aiohttp and azure-identity are well-known libraries, azure-ai-voicelive is a community-provided package that serves as the core SDK for this skill's functionality.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its processing of real-time audio and text data from external sources.\n
  • Ingestion points: Untrusted data enters the agent context through conn.input_audio_buffer and the event.transcript attribute in the Quick Start and Event Handling sections of SKILL.md.\n
  • Boundary markers: The provided code snippets do not include explicit boundary markers or instructions to the model to ignore potential commands embedded within the audio or transcript data.\n
  • Capability inventory: The skill has the capability to execute function calls (handle_function), update session configurations, and create or cancel responses.\n
  • Sanitization: There is no evidence of sanitization, filtering, or validation of the incoming audio stream or transcribed text before it is processed by the AI model.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 20, 2026, 08:58 AM