The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the belt CLI and various sub-skills from the inference.sh platform. It provides a link to installation instructions hosted on a vendor-related GitHub repository (https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including text prompts and media URLs, which are passed to downstream AI models. This creates a surface for indirect prompt injection if the processed content contains instructions designed to manipulate the model's output or the agent's behavior.
Ingestion points: JSON input fields prompt, image_url, audio_url, video_url, and reference_image used in belt app run commands.
Boundary markers: No explicit delimiters or warnings for the agent to ignore instructions within the processed data are specified in the instructions.
Capability inventory: The skill has access to the Bash tool to execute belt CLI commands, which interact with external APIs.
Sanitization: The instructions do not define any filtering or sanitization of the external content before it is processed.

ai-video-generation