The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing official Remotion ecosystem packages (e.g., @remotion/media, @remotion/captions, @remotion/three) and utilities through standard package managers (npm, yarn, pnpm, bun).
[COMMAND_EXECUTION]: Provides implementation examples for audio transcription that use child_process.execSync to run FFmpeg, which is a standard procedure for media processing within the Remotion development environment.
[CREDENTIALS_UNSAFE]: Encourages the secure handling of API keys (such as Mapbox access tokens) by using .env files, following industry standard security practices for local development and preventing accidental exposure of secrets.
[PROMPT_INJECTION]: The skill includes guidance on processing external data (e.g., Lottie JSON or dynamic metadata via fetch), creating a potential surface for indirect prompt injection.
Ingestion points: calculateMetadata (remote data URLs), display-captions.md (external JSON), lottie.md (remote assets).
Boundary markers: Delimiters are absent in the examples provided.
Capability inventory: Utilizes fetch for data retrieval and child_process.execSync in user-facing scripts.
Sanitization: Specifically recommends implementing Zod schemas in rules/parameters.md to validate and sanitize dynamic data.
[SAFE]: The content is purely instructional, focusing on best practices for video creation and promoting safe patterns like schema validation and proper resource management.

remotion-best-practices