The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the uv run command to execute localized CLI tools including comfy-videogen, comfy-media, and comfy-model-downloader. These commands are used for generating video content, managing local media galleries, and initializing model checkpoints.
[EXTERNAL_DOWNLOADS]: The instructions reference comfy-model-downloader to fetch required model weights (e.g., LTX 2.3 variants) to the /mnt/models/comfyui directory. This is presented as a standard setup procedure for local inference.
[PROMPT_INJECTION]: The skill accepts user-defined text strings via the --prompt flag for video generation. This constitutes a standard input surface for indirect prompt injection, which is inherent to the skill's primary purpose of text-to-video synthesis.

comfy-videogen