The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to install various official and community packages from the Remotion ecosystem. Examples include commands like npx remotion add @remotion/three, @remotion/media, and @remotion/captions using common package managers (npm, bun, yarn, pnpm).\n- [EXTERNAL_DOWNLOADS]: The instructions direct the agent to fetch media assets (images, videos, audio), Lottie animations, and subtitle files from external sources such as LottieFiles and official Remotion media test sites. It also suggests using browsing tools to fetch documentation and guidelines from remotion.dev. These references are to well-known and reputable services.\n- [INDIRECT_PROMPT_INJECTION]: The skill presents an indirect prompt injection attack surface by processing untrusted external data.\n
Ingestion points: Data enters the agent's context through props.dataUrl and fetch operations in the calculateMetadata function (rules/calculate-metadata.md), as well as through the extraction of brand details from user-provided URLs in Step 0 of the creation process (SKILL.md).\n
Boundary markers: There are no explicit instructions or boundary markers defined to ensure that the agent ignores potential natural language instructions embedded within the fetched external content or brand guidelines.\n
Capability inventory: The agent is empowered to perform network requests, file system modifications within the project, and command execution for package management.\n
Sanitization: The skill does not provide methods or instructions for sanitizing or escaping the data retrieved from external URLs before it is interpolated into the composition's props or used to drive logic.

remotion