voice-agents

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

• Third-party content exposure detected (medium risk: 0.30). In both the OpenAI Realtime API and the pipeline example, the agent ingests outsider-authored free text produced from the live caller’s audio transcript (e.g., event.transcript / userText) and sends it directly as the LLM user message without requiring selection of a specific pre-published item.