The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external media assets (images, videos, 3D models) which can serve as a vector for indirect prompt injection via embedded metadata or hidden instructions.
Ingestion points: External URLs (Appendix A) and user-provided reference files stored in '03-assets/refs/'.
Boundary markers: None present; the skill treats content from sourced assets as design inputs without explicit safety boundaries.
Capability inventory: Includes 'ffmpeg' for media processing, 'npx' for utility execution, and network access for asset retrieval.
Sanitization: No evidence of metadata stripping or content validation before processing.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various command-line tools to process assets.
Executes 'ffmpeg' for frame extraction and media conforming.
Uses 'npx' to run 'hyperframes' utilities for background removal, transcription, and text-to-speech.
Suggests local execution of tools like 'rembg' and 'ComfyUI' for asset generation.
[EXTERNAL_DOWNLOADS]: The skill provides a comprehensive list of external services and repositories for sourcing creative assets.
Downloads 3D models and textures from Sketchfab, Poly Haven, and NASA.
Fetches tools and models from GitHub organizations including Microsoft, OpenAI, and various open-source projects.
Interfaces with numerous AI generation platforms like Midjourney, ElevenLabs, and Runway.

motion-assets