twilio-inbound-voice
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data via real-time audio streams, which are transcribed into text for the agent to process.
- Ingestion points: Real-time audio received through
WebSocket /webhooks/twilio/media-streamand transcribed by the Whisper STT service. - Boundary markers: The documentation does not specify explicit boundary delimiters for the transcribed text before it is passed to the LLM.
- Capability inventory: The skill has the ability to execute network queries, status checks, and action commands (e.g., restarting monitoring agents, clearing counters) across connected MCP servers.
- Sanitization: The skill implements a verbal confirmation flow ('Confirmation Flow') requiring the user to explicitly say 'yes' before any impactful action is executed. Additionally, response content is sanitized to remove sensitive information like IPs or credentials before conversion to speech.
- [EXTERNAL_DOWNLOADS]: The skill architecture integrates with well-known external service providers for core processing tasks.
- Evidence: The workflow utilizes the OpenAI Whisper API for speech-to-text transcription and Twilio services for media streaming and text-to-speech generation.
Audit Metadata