elevenlabs-agents

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

• Third-party content exposure detected (high risk: 0.95). The required runtime workflow for this skill is a voice-agent conversation where the agent ingests the OUTSIDER-authored user’s spoken text (transcript) into the LLM context via the ElevenLabs conversation/ASR pipeline (user messages are not authored by the operating user).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The widget embed HTML includes a runtime script tag loading and executing remote JavaScript (https://elevenlabs.io/convai-widget/index.js), which runs in the client and directly implements/controls the agent widget behavior.