together-audio

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill transcribes audio data which can contain embedded linguistic instructions, creating an indirect prompt injection surface for agents consuming the output.\n
  • Ingestion points: Audio data is ingested via local file system paths and remote URLs in scripts/stt_transcribe.py, scripts/stt_transcribe.ts, and references/stt-models.md.\n
  • Boundary markers: The scripts and documentation do not implement delimiters or specific instructions to the agent to treat the transcribed text as untrusted data.\n
  • Capability inventory: The skill utilizes file system access (reading audio files, writing generated speech) and network access to Together AI's infrastructure (api.together.ai and api.together.xyz).\n
  • Sanitization: The transcription process does not include filtering or sanitization of the resulting text before it enters the agent's context.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 08:06 PM