The Agent Skills Directory

[SAFE]: The skill instructions and associated commands were analyzed, and no malicious patterns, obfuscation techniques, or unauthorized security-sensitive operations were found.
[COMMAND_EXECUTION]: The skill utilizes local CLI tools (comfy-videogen, comfy-models) via the standard uv runner to perform video generation and manage models. These operations are restricted to the intended purpose of the skill and do not involve unauthorized shell execution.
[EXTERNAL_DOWNLOADS]: The skill describes downloading model checkpoints from Lightricks, a well-known organization in the AI space, using a project-specific downloader. These downloads target official model weights and do not include untrusted or unverifiable scripts.
[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection as it processes user-supplied prompts for video generation. However, these prompts are consumed by the LTX 2.3 generation model as data and do not influence the agent's control flow or execution logic.
Ingestion points: User-supplied --prompt string, image, and video files (SKILL.md).
Boundary markers: None explicitly defined in the CLI command examples.
Capability inventory: Local subprocess execution of the comfy-videogen CLI.
Sanitization: Not explicitly defined within the skill documentation; relies on the underlying CLI tool's handling of arguments.

comfy-motion-track-control