The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted video URLs and user-supplied prompts which are passed to the video processing model, creating a surface for indirect prompt injection.
Ingestion points: The skill accepts video_url and prompt parameters within a JSON payload for the runcomfy command as seen in SKILL.md.
Boundary markers: The skill specifies that inputs are passed via a JSON string to the --input flag, intended to prevent shell expansion, and includes warnings for agents to treat external content as untrusted.
Capability inventory: The skill uses the Bash(runcomfy *) tool to execute video generation and editing commands.
Sanitization: Documentation advises agents to monitor for output divergence and to only process URLs explicitly provided by the user.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the @runcomfy/cli package from the NPM registry. This package is a verified resource provided by the skill author (gencraft-labs).
[COMMAND_EXECUTION]: The skill utilizes the runcomfy CLI to perform video outpainting operations and handle user authentication via the runcomfy login command.

video-outpainting