The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references external resources for tool installation and documentation, primarily targeting inference.sh and its official GitHub repository (inference-sh/skills). These are consistent with the skill's purpose as a wrapper for a specific service.
[COMMAND_EXECUTION]: The skill requires the use of the belt CLI tool to interact with the API. It provides commands for logging in, searching for models, and running video generation jobs. These commands are standard for CLI-based service integration.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided prompts which are then sent to external AI models. While this presents an attack surface where malicious input could influence model output, it is the primary intended function of the skill and incorporates standard JSON structures for input passing.
Ingestion points: User-defined prompts passed to the --input argument in SKILL.md examples.
Boundary markers: Prompts are encapsulated within JSON strings (e.g., {"prompt": "..."}).
Capability inventory: The skill uses the belt CLI via the Bash tool to perform network-based video generation.
Sanitization: Standard CLI argument passing is used; model-level safety filters from Google Veo are expected to handle content generation risks.

google-veo