deepgram-js-audio-intelligence

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly shows ingesting remote audio via transcribeUrl (see "Quick start — REST with analytics" example) and WebSocket connections, which fetch arbitrary public URLs and return summaries/topics/intents that the agent is expected to read and could materially influence subsequent actions.