The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is configured to execute the runcomfy CLI tool using the Bash environment. The skill specifically restricts its capabilities to subcommands of runcomfy, which adheres to the principle of least privilege.
[EXTERNAL_DOWNLOADS]: The documentation provides standard instructions for installing the necessary command-line interface via npm or npx from the official @runcomfy/cli package. It explicitly warns against unsafe installation methods like piping remote scripts directly into the shell.
[PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection when processing reference image URLs or using web search features. The author has included specific mitigation strategies for the agent, such as prioritizing user instructions over content found in external assets.
[CREDENTIALS_UNSAFE]: The skill documents the use of an API token for authentication. It recommends using an environment variable (RUNCOMFY_TOKEN) or an interactive login that stores the token with restricted file permissions (0600), which is an industry-standard practice for CLI tools.

ai-image-generation