The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute belt commands for logging in, listing available AI models, and running video generation tasks.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill references and encourages the installation of external components, including the belt CLI tool (npx skills add belt-sh/cli) and additional model-specific skills from the inference-sh GitHub repository. These dependencies are consistent with the primary purpose of the skill.
[DATA_EXPOSURE_AND_EXFILTRATION]: Execution of belt app run commands involves sending user-provided prompts and media URLs to the inference.sh API for processing. This is the intended functionality of the service.
[INDIRECT_PROMPT_INJECTION]: The skill processes external media via URLs (images, audio, video) which are passed to AI models, creating a potential surface for indirect prompt injection.
Ingestion points: Media URLs provided in the --input argument of belt app run commands in SKILL.md.
Boundary markers: None explicitly defined in the CLI command structure for external data.
Capability inventory: Uses Bash(belt *) to execute remote AI inference.
Sanitization: Not explicitly defined within the skill; relies on the underlying belt CLI tool and the model provider's safety filters.

ai-video-generation