The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data such as quote text, script beats, and captions, which are then interpolated into JSON property files and used as inputs for video rendering.
Ingestion points: User input provided via triggers for quote cards, clips, and audiograms (SKILL.md).
Boundary markers: The skill does not explicitly use delimiters or warnings to ignore instructions embedded in user-provided text, though it uses structured JSON for most props.
Capability inventory: The skill utilizes Bash, Write, Edit, and Read tools to manage files and execute rendering pipelines.
Sanitization: Some inputs are processed via JSON.stringify within Node.js snippets, reducing the risk of shell injection during transcription or check phases.
[COMMAND_EXECUTION]: The instructions involve extensive use of shell commands including ffmpeg, ffprobe, npx remotion, and a vendor-specific video-production CLI to manage the video lifecycle.
[DYNAMIC_EXECUTION]: The skill uses node -e (eval) to run inline JavaScript/TypeScript snippets for transcription (using Whisper), slideshow risk assessments, and post-render frame analysis. These snippets are provided as static templates within the instructions and operate on local project data.

video-render