The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes patterns for fetching and processing external data that could potentially contain malicious instructions intended to influence the agent's behavior.
Ingestion points: The calculateMetadata function in rules/calculate-metadata.md fetches data from a caller-provided props.dataUrl. Additionally, rules/import-srt-captions.md fetches and parses external .srt subtitle files.
Boundary markers: The provided code snippets do not include explicit delimiters or instructions to the agent to ignore potentially embedded commands within the fetched data.
Capability inventory: The skill utilizes fetch for network operations and various parsing utilities (parseSrt, json()) to integrate external content into the video rendering pipeline.
Sanitization: The examples do not demonstrate sanitization or validation of the remote content before it is transformed into component props or rendered.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install several external Node.js packages to enable specific functionalities.
The skill recommends installing official framework modules including @remotion/three, @remotion/media, @remotion/captions, @remotion/google-fonts, @remotion/fonts, @remotion/gif, @remotion/layout-utils, @remotion/lottie, and @remotion/transitions.
It also introduces a dependency on mediabunny, a third-party library used for media metadata extraction and frame processing across multiple rule files (e.g., rules/can-decode.md, rules/extract-frames.md).
It fetches Lottie assets from assets4.lottiefiles.com and documentation from remotion.dev.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands to manage the project environment.
Evidence: Multiple files provide commands like npx remotion add [package] to add dependencies via npm, bun, yarn, or pnpm.

remotion-video-creation