telnyx-voice-streaming-python
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by processing real-time voice transcriptions. If the resulting text is fed into an LLM without appropriate safeguards, an attacker could potentially influence the agent's behavior via spoken instructions.
- Ingestion points: The
transcriptionwebhook event described inSKILL.mdprovides external, untrusted text data to the system. - Boundary markers: Absent; there are no instructions or delimiters defined to help the agent distinguish between its own logic and the transcribed data.
- Capability inventory: The skill utilizes the
telnyxSDK to perform call control actions likestart_forking,start_streaming, andstart_transcription. - Sanitization: Absent; the example implementation does not include logic for filtering or escaping the transcribed content before use.
Audit Metadata