The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for fetching external data at runtime to parameterize video compositions. This represents an attack surface where maliciously crafted external data could attempt to influence the agent's behavior during the video generation process.
Ingestion points: Data is ingested from external URLs in calculate-metadata.md (via props.dataUrl) and lottie.md (via fetch for Lottie JSON assets).
Boundary markers: The instructions do not specify any boundary markers or delimiters to isolate untrusted external data from the agent's instructional context.
Capability inventory: The skill requires the ability to execute shell commands using npx remotion for tasks such as starting the studio, rendering frames, and adding packages.
Sanitization: The provided examples do not demonstrate sanitization or validation of the external JSON data before it is used to populate component props.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from well-known external services to support its functionality.
Evidence: CesiumFlythrough.tsx dynamically loads the CesiumJS library from cesium.com at runtime. Other components reference media assets, map tiles, and fonts from well-known providers including remotion.media, googleapis.com, maptiler.com, and mapbox.com.

remotion-markup