remotion-best-practices
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides numerous shell command snippets for installing official Remotion extensions (e.g., @remotion/three, @remotion/media, @remotion/lottie) using common package managers like npm, yarn, bun, and pnpm.
- [EXTERNAL_DOWNLOADS]: Guidelines include code for fetching remote assets such as videos, images, and Lottie animations from well-known domains like lottiefiles.com and remotion.media.
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch additional technical instructions from the official Remotion documentation at remotion.dev using WebFetch.
- [PROMPT_INJECTION]: The skill documents patterns for ingesting untrusted external data, such as .srt subtitle files and JSON metadata via fetch, which constitutes a surface for indirect prompt injection.
- Ingestion points: calculate-metadata.md (reads from props.dataUrl), import-srt-captions.md (fetches .srt files), lottie.md (fetches JSON animation data).
- Boundary markers: None are specified in the code examples for isolating the external content from the prompt logic.
- Capability inventory: The documented capabilities are limited to network requests, file reading (staticFile), and UI rendering via React; no subprocess execution or dynamic code evaluation of the ingested data is present.
- Sanitization: The examples show standard JSON and text parsing but do not include specific sanitization or validation logic for the retrieved strings.
Audit Metadata