The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md instructions for executing the generation tool use string interpolation to place the generated prompt directly into a shell command (--prompt "[构建好的英文提示词]"). This pattern creates a risk of command injection if the prompt contains shell-active characters like double quotes, backticks, or semicolons that are not properly escaped.
[CREDENTIALS_UNSAFE]: The skill requires the agent to ask the user for an API key and store it in a plaintext configuration file at ~/.config/ima2gen/config.json. While this is common for local developer tools, it involves the handling and local persistence of sensitive credentials.
[EXTERNAL_DOWNLOADS]: The Python execution script (cli/ima2gen.py) makes outbound network requests to several third-party API providers (including api.apimart.ai, code.rustnest.cc, and bmai.kun8.vip) to generate images and subsequently downloads the resulting image files from the URLs returned by these services.
[PROMPT_INJECTION]: The skill ingests untrusted user descriptions to construct professional image prompts. The instructions lack boundary markers or sanitization steps before the generated prompt is passed into a shell command, which provides a surface for indirect prompt injection to influence the host environment via command arguments.

ima2gen