remotion

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

• Third-party content exposure detected (high risk: 0.90). The skill includes explicit runtime fetches of arbitrary remote URLs (e.g., rules/calculate-metadata.md shows calculateMetadata calling fetch(props.dataUrl) and using the returned JSON to set duration/width/props), so untrusted third-party content is ingested and can directly change composition behavior and decisions.

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

• Potentially malicious external URL detected (high risk: 0.70). The transcribe-captions example calls installWhisperCpp/downloadWhisperModel/transcribe from @remotion/install-whisper-cpp which (per the linked docs) downloads and runs whisper.cpp and model files at runtime (see https://www.remotion.dev/docs/install-whisper-cpp), so remote code/models are fetched and executed as a required dependency.