The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's functionality is centered around the runcomfy CLI tool. It utilizes the allowed-tools frontmatter field to restrict the agent's shell access specifically to the runcomfy command, effectively minimizing the attack surface.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the official @runcomfy/cli via NPM. This is a standard installation procedure from a well-known service associated with the skill's stated purpose. The documentation explicitly warns against piping arbitrary remote scripts into a shell, which is a key security recommendation.
[PROMPT_INJECTION]: The skill identifies the risk of indirect prompt injection via user-supplied media URLs (images, video, audio). It provides clear instructions for the agent to validate user-provided inputs and refuse requests involving non-consensual or harmful content generation, such as defamatory or sexually explicit deepfakes.
[CREDENTIALS_UNSAFE]: The skill describes safe credential management by pointing to the CLI's default configuration path (~/.config/runcomfy/token.json) with secure file permissions (0600) and suggesting the use of environment variables for automation, avoiding hardcoded secrets.

face-swap