twilio-voice-conversation-relay
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user speech transcripts (
voicePrompt) received via WebSockets to drive LLM responses. This creates an attack surface where a caller could attempt to inject malicious instructions into the LLM prompt via speech-to-text. While the risk is present in any LLM-based voice system, the skill relies on the implementer's LLM guardrails. - [SAFE]: The skill references and uses well-known Twilio services and libraries (
twilioNPM package,websocketsPython package). These are documented neutrally as they are trusted components of the Twilio ecosystem.
Audit Metadata