The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute subcommands of the runcomfy CLI tool to log in and generate videos, as specified in the allowed-tools configuration.
[EXTERNAL_DOWNLOADS]: The skill references the installation of the @runcomfy/cli package via npm, which is the official tool for the service integrated by the skill.
[CREDENTIALS_UNSAFE]: The documentation describes the storage of API tokens in ~/.config/runcomfy/token.json and the use of the RUNCOMFY_TOKEN environment variable, which are standard but sensitive credential management locations for the underlying CLI.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external media assets (images and audio files) provided by users, which can serve as vectors for indirect prompt injection.
Ingestion points: The image_url, audio_url, and prompt fields within the JSON input for the runcomfy run command in SKILL.md.
Boundary markers: Inputs are structured using JSON to prevent command injection at the CLI level, though the skill does not define specific text-based delimiters for the prompts processed by the backend AI models.
Capability inventory: The runcomfy CLI has capabilities for authenticated network communication with the RunComfy API and file system access for downloading output files.
Sanitization: The instructions explicitly direct the agent to only process user-provided URLs and provide guidance on detecting unexpected generation behavior that might indicate poisoned assets.

ai-avatar-video