The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines patterns that introduce an indirect prompt injection surface. Specifically, it encourages fetching external data at runtime to influence rendering and metadata.
Ingestion points: rules/calculate-metadata.md (data from props.dataUrl), rules/lottie.md (external JSON assets), rules/display-captions.md (external caption files).
Boundary markers: No explicit instruction delimiters or 'ignore embedded instruction' warnings are used in the provided examples.
Capability inventory: The skill utilizes subprocess execution via npx remotion, ffmpeg, and Node.js scripts with file system and network access.
Sanitization: No explicit validation or sanitization logic is provided for the fetched external content before it is interpolated into the composition logic.
[EXTERNAL_DOWNLOADS]: The skill recommends installing several official Remotion ecosystem packages and tools (e.g., @remotion/media, @remotion/transitions, @remotion/google-fonts). It also uses @remotion/install-whisper-cpp to download the Whisper.cpp binary and pre-trained models, which is a trusted framework utility.
[COMMAND_EXECUTION]: The documentation provides numerous instructions for executing shell commands, including package management (npx, bunx, yarn, pnpm), framework-specific CLI tools, and media processing utilities like ffmpeg and ffprobe.

remotion-best-practices