The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references installation guidelines hosted on the official inference-sh GitHub repository for the required command-line tools.
[COMMAND_EXECUTION]: The skill utilizes the belt command-line interface to interact with video generation models. This execution is restricted via the allowed-tools configuration to only allow belt commands, which is a recommended security practice.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external data by accepting image and video URLs as inputs for model inference.
Ingestion points: External URLs for images and videos are passed as arguments to the belt command in SKILL.md.
Boundary markers: Not explicitly defined in the templates, but usage is restricted to structured JSON inputs for the CLI tool.
Capability inventory: Uses the belt CLI for remote API interaction (inference.sh).
Sanitization: Relies on the underlying belt CLI and the inference.sh platform to handle the provided URLs safely.

happyhorse