The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of the runcomfy CLI tool to submit requests to the Model API and manage local file downloads of generated images.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the @runcomfy/cli package from the NPM registry to function.
[CREDENTIALS_UNSAFE]: The skill manages a sensitive API token, which it stores in ~/.config/runcomfy/token.json or retrieves from the RUNCOMFY_TOKEN environment variable for authentication with the RunComfy platform.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection due to how it handles untrusted external data.
Ingestion points: The skill ingests user-provided text prompts and external image URLs (in the edit endpoint) which are processed by the image generation model.
Boundary markers: While the skill notes that prompts are passed as JSON to the CLI to prevent shell injection, it lacks delimiters or specific instructions to the model to ignore potentially malicious directives embedded within those inputs.
Capability inventory: The skill has the capability to write files to the local file system (via the --output-dir flag) and perform network operations to the RunComfy API.
Sanitization: There is no evidence of sanitization or filtering performed on the content of the prompts or the data fetched from external URLs before it is sent to the model.

gpt-image-2