The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and execute external binaries and fetch remote assets.
rules/transcribe-captions.md provides a Node.js script using the @remotion/install-whisper-cpp package to download and install the whisper.cpp binary (version 1.5.5) to the project directory.
rules/lottie.md fetches animation data from https://assets4.lottiefiles.com/packages/lf20_zyquagfl.json.
rules/voiceover.md fetches audio from the ElevenLabs API (https://api.elevenlabs.io/v1/text-to-speech/).
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) by processing untrusted data.
Ingestion points: rules/calculate-metadata.md fetches JSON from a user-provided props.dataUrl. rules/compositions.md fetches video metadata from an external API based on props.videoId. rules/lottie.md and rules/display-captions.md fetch and parse external JSON content.
Boundary markers: No specific boundary markers or "ignore embedded instructions" warnings are used for content fetched from remote URLs.
Capability inventory: The skill provides templates for writing files to the local disk (fs.writeFileSync in rules/transcribe-captions.md and rules/voiceover.md) and executing CLI-based video rendering commands.
Sanitization: While rules/parameters.md introduces zod for prop schema validation, the examples for fetching external data do not include validation of the fetched payload's structure or content before use in the video composition.

remotion