The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the runcomfy CLI binary to perform video generation tasks (e.g., runcomfy run kling/kling-3-0/...). This is the primary method of interaction and is used to pass parameters to the backend service.
[EXTERNAL_DOWNLOADS]: The instructions direct users to install the @runcomfy/cli package from the npm registry. Additionally, the CLI automatically downloads generated video assets from *.runcomfy.net and *.runcomfy.com domains upon completion of a task.
[CREDENTIALS_UNSAFE]: The skill manages API authentication using the RUNCOMFY_TOKEN environment variable or by storing a persistent token in ~/.config/runcomfy/token.json following a browser-based login flow. This is a standard credential management pattern for CLI tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and external image URLs (image_url) which are passed to the remote video generation model.
Ingestion points: User input fields prompt, image_url, and tail_image_url defined in the input schema.
Boundary markers: Arguments are encapsulated within a JSON structure passed via the --input flag to the CLI, providing a data/instruction boundary.
Capability inventory: The skill executes shell commands via the runcomfy CLI and performs network-based file downloads to the local file system.
Sanitization: The skill relies on the CLI's JSON parsing and the backend model's safety filters to handle potentially malicious content within images or text prompts.

kling-3-0