The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the installation of custom nodes by cloning repositories from the internet using the install_custom_nodes tool and manual installation instructions. This allows for the execution of arbitrary code within the local environment.
[EXTERNAL_DOWNLOADS]: The skill enables downloading files from external URLs through the download_model and download_civitai_model tools. These tools accept arbitrary URLs as arguments.
[COMMAND_EXECUTION]: The skill uses npx to execute the MCP server and provides tools to manage the ComfyUI lifecycle, including restarting and stopping the service via shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of generation prompts and workflow definitions which could be used to trick the agent into performing dangerous actions.
Ingestion points: The prompt and workflow parameters in tools like generate_image and enqueue_workflow (SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the documentation.
Capability inventory: The skill possesses high-privilege capabilities including Git cloning, file downloading, and system process management.
Sanitization: No sanitization or validation of input content is documented.

comfyui-mcp-agent