voice-ai-development

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). This skill explicitly ingests untrusted user-generated content via Vapi webhooks (e.g., event["transcript"] in the Vapi webhook) and Deepgram real-time transcription (on_transcript handler) and then processes those transcripts to trigger function calls and workflow actions, so third-party content can influence agent behavior.