The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references installation guides and documentation from the inference.sh GitHub repository and domain. These are standard references for configuring the required environment for the skill's functionality.
[COMMAND_EXECUTION]: The skill leverages the belt CLI via the Bash tool to interact with the inference service. The allowed-tools configuration appropriately limits the tool's scope to the belt command.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input, such as text prompts and media URLs, which are interpolated into CLI commands for video processing.
Ingestion points: User-provided prompts, image URLs, and video URLs (SKILL.md).
Boundary markers: None explicitly present in the prompt templates.
Capability inventory: Bash(belt *) tool access (SKILL.md).
Sanitization: The skill relies on the belt CLI and the downstream AI model's safety guardrails to handle potentially malicious inputs.

happyhorse