deepgram-python-audio-intelligence

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly ingests audio from arbitrary URLs or streams (e.g., "client.listen.v1.media.transcribe_url" and the "You have audio (file, URL, or live stream)" guidance in SKILL.md), causing the agent to read and analyze untrusted third‑party content (audio) whose inferred intents/summaries could influence downstream decisions, so it can enable indirect prompt injection.