The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of local utilities for media processing, including FFmpeg for audio/video manipulation and the official HyperFrames and HeyGen CLI tools for resource resolution. All commands are executed using safe patterns like argument lists to prevent shell injection.
[EXTERNAL_DOWNLOADS]: Media assets (music, SFX, images) are fetched from trusted vendor domains (heygen.ai, googleapis.com, elevenlabs.io) and well-known platforms like Openverse (via api.openverse.org). The skill implements a robust SSRF protection layer in scripts/lib/media-fetch.mjs that blocks requests to local and private IP ranges.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as video transcripts and composition metadata, which are interpolated into prompts for AI-backed media generation and TTS services. While this presents a potential risk surface common in AI agents, the skill includes explicit documentation and instructions for agents to perform quality checks and filtering on the resulting content.
[DYNAMIC_EXECUTION]: To support local media generation pathways (e.g., MusicGen), the skill automatically installs necessary Python dependencies from the official PyPI registry when those specific features are triggered. The installed packages (e.g., transformers, torch, google-genai) are well-known machine learning libraries required for the skill's core functionality.

media-use