agents

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

• Third-party content exposure detected (medium risk: 0.30). ElevenLabs agents ingest outsider-authored free text from real-time conversation user input (e.g., via the widget/Conversation.startSession flow that passes user transcript content to the agent/LLM).

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

• Potentially malicious external URL detected (high risk: 0.90). The embed script https://unpkg.com/@elevenlabs/convai-widget-embed is loaded at runtime (script src in SKILL.md and widget docs) and will execute remote JavaScript in the client, so it is a runtime external dependency that fetches and runs code.