The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and execute external, untrusted data in the form of community ComfyUI recipes through workflow_json and workflow_path parameters.
Ingestion points: User-provided workflow_json and workflow_path, as well as model download URLs defined in the missing_models structured payload.
Boundary markers: There are no explicit instructions for the agent to validate the structure or content of the workflow JSON against a schema before submission.
Capability inventory: The skill enables network communication (sending prompts to the COMFYUI_SERVER_URL) and file system access (writing models to ComfyUI/models/ subdirectories).
Sanitization: No sanitization of the external JSON content or validation of model download URLs is performed within the skill instructions.
[EXTERNAL_DOWNLOADS]: The skill facilitates the downloading of external assets by instructing the agent to use URLs provided in the data.missing_models[] list to resolve dependency issues.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to perform file system operations, such as placing downloaded models and LoRAs into specific directory structures (e.g., ComfyUI/models/diffusion_models/), and suggests interacting with the ComfyUI Manager for node installation.

comfyui