The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill relies on the @runcomfy/cli tool, which is retrieved from the official npm registry. This external dependency is essential for the skill's primary function of interfacing with RunComfy's video models.
[COMMAND_EXECUTION]: The skill invokes the runcomfy command-line utility to perform video processing. This usage is appropriately restricted to the specific command via the platform's tool access controls.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, specifically video URLs and user-supplied prompts, which could be used to influence the agent's behavior.
Ingestion points: The video_url and prompt parameters passed to the runcomfy run command in SKILL.md are the primary entry points for external data.
Boundary markers: Input data is encapsulated within a JSON object, providing a layer of separation from the command shell.
Capability inventory: The runcomfy tool can access network resources to process video and write output files to the local system.
Sanitization: The skill includes explicit security guidance, instructing the agent to only ingest content from user-verified sources and to monitor for unexpected outputs that might indicate a compromised source.

video-outpainting