The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents several methods for ingesting external data into video compositions, creating a potential surface for indirect injection.
Ingestion points: The getInputProps() function in reference.md allows data ingestion via CLI flags, and the calculateMetadata function in reference.md demonstrates fetching data from external API endpoints.
Boundary markers: The documentation does not provide examples of delimiters or warnings to prevent the agent from following instructions embedded in the ingested data.
Capability inventory: The skill documents powerful tools in reference.md including renderMedia (video rendering), deployFunction (AWS Lambda deployment), and deploySite (site deployment to cloud providers).
Sanitization: There is no evidence of sanitization, escaping, or schema validation provided for external data before it is processed by the rendering engine.
[DYNAMIC_EXECUTION]: The skill provides technical documentation for Remotion's core rendering and deployment capabilities which involve runtime code compilation and execution.
Runtime compilation: reference.md describes the use of @remotion/bundler's bundle() function to compile source code into executable bundles at runtime.
Dynamic loading: reference.md details the execution of bundled code using @remotion/renderer functions such as renderMedia and renderStill.
Remote deployment: The skill documents the use of @remotion/lambda to deploy and execute code on remote AWS infrastructure.

remotion