deepgram-core-workflow-a

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly accepts and fetches remote, arbitrary URLs for transcription (see "Audio files or URLs to transcribe" in SKILL.md and the transcribeUrl/transcribe_url methods in services/transcription.ts and services/transcription.py), which means untrusted third‑party audio content could be ingested and its transcribed instructions could influence subsequent actions.