remotion-best-practices

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

• Third-party content exposure detected (high risk: 0.80). The calculate-metadata rule (rules/calculate-metadata.md) explicitly shows fetching arbitrary JSON via fetch(props.dataUrl) and using that data to set composition props/duration, meaning the skill instructs ingesting untrusted third‑party URLs (props.dataUrl) whose contents directly affect runtime decisions.

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

• Potentially malicious external URL detected (high risk: 0.80). The transcribe-captions rule invokes installWhisperCpp() and downloadWhisperModel() (see https://www.remotion.dev/docs/install-whisper-cpp), which at runtime download and install remote whisper.cpp binaries/models (fetching and executing external code) and are required for the skill to work.