voice-isolator

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md "Isolating from a URL" example explicitly fetches audio via requests.get(...) from arbitrary URLs (e.g., https://example.com/noisy.mp3), and the skill's "Common Workflows" directs feeding isolated audio into speech-to-text, so untrusted third-party audio can be ingested and then influence downstream transcription-driven actions.