soniox

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs ingesting audio from public HTTP(S) URLs (see references/async-api.md "audio_url: Public HTTP(S) URL to the audio file") and includes SDK examples that stream arbitrary external URLs (e.g., references/sdks.md streaming "https://soniox.com/media/examples/coffee_shop.mp3" and web SDK examples using audioElement.src = "https://example.com/audio.mp3"), so untrusted third-party content would be transcribed and could indirectly inject instructions that influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes an explicit runtime import of remote executable JS from the CDN URL https://unpkg.com/@soniox/speech-to-text-web?module which will be fetched and executed in clients (i.e., runs remote code as a required SDK dependency), creating a high-confidence runtime external dependency.