agents

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

• Third-party content exposure detected (medium risk: 0.30). In the ElevenLabs agents conversation runtime (widget/client session), the agent ingests outsider-authored free text from the user’s speech/transcript via onUserTranscript/ASR, which the LLM then uses for prompting.

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

• Potentially malicious external URL detected (high risk: 0.90). The page embeds the remote widget script https://unpkg.com/@elevenlabs/convai-widget-embed at runtime (script src), which downloads and executes third-party JavaScript required for the widget, so it is a runtime external dependency that executes remote code.