remotion-best-practices

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

• Third-party content exposure detected (high risk: 0.80). The skill includes explicit runtime fetches of public third‑party content (e.g., SKILL.md and rules/calculate-metadata.md show fetch(https://api.example.com/video/${props.videoId}) and fetch(props.dataUrl), rules/lottie.md fetches a Lottie JSON from assets4.lottiefiles.com, and voiceover.md calls the ElevenLabs API), and that fetched, untrusted content is parsed and used to set composition metadata, props, or assets — so external data can materially influence the agent's decisions and workflow.

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

• Potentially malicious external URL detected (high risk: 0.90). The transcribe-captions.md example invokes @remotion/install-whisper-cpp (see https://www.remotion.dev/docs/install-whisper-cpp) and calls installWhisperCpp/downloadWhisperModel in a runtime script, which downloads and installs/executes remote whisper.cpp binaries and models needed for transcription, so it fetches and runs external code at runtime.