The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the fetching of data from external, potentially untrusted URLs to drive video composition properties and metadata.
Ingestion points: Network operations in rules/calculate-metadata.md, rules/compositions.md, and rules/lottie.md ingest data via fetch() from sources like props.dataUrl or external asset hosts.
Boundary markers: The instructions do not define boundary markers or instruct the agent to sanitize or ignore instructions embedded within the fetched data.
Capability inventory: The skill uses network requests (fetch), file metadata extraction (mediabunny), and rendering of dynamic React components.
Sanitization: Fetched JSON data is directly spread into component props or used for duration calculations without validation or filtering in the provided examples.
[EXTERNAL_DOWNLOADS]: The skill contains references to external services for assets and documentation, such as assets4.lottiefiles.com for Lottie animations and official remotion.dev documentation links.
[COMMAND_EXECUTION]: The rule files include standard shell commands for project setup and package installation (e.g., npx remotion add) targeting framework-specific libraries like @remotion/three, @remotion/media, and @remotion/captions.

remotion-video-creation