The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external image data provided via HTTPS URLs. This creates a surface for indirect prompt injection where an attacker could embed malicious instructions within an image to influence the model's behavior.
Ingestion points: The images parameter in the runcomfy run command accepts an array of up to 10 external URLs which are fetched by the model server.
Boundary markers: Input is structured as a JSON object passed to the CLI via the --input flag, providing structural separation between user instructions and external data URLs.
Capability inventory: The skill executes the runcomfy CLI, which performs network operations to the runcomfy.net API and writes output files to the local filesystem.
Sanitization: No specific image-content sanitization or filtering is performed by the skill; protection against malicious content relies on the model provider's internal safety guardrails.
[COMMAND_EXECUTION]: The skill invokes the runcomfy CLI to communicate with the Model API. This is the intended primary function of the skill and involves standard command-line argument passing.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs the user to install the @runcomfy/cli package from the npm registry and utilizes vendor-owned domains for API interaction. These are established vendor resources necessary for the skill's functionality.

gpt-image-edit