The Agent Skills Directory

[COMMAND_EXECUTION]: The Bash scripts provided in SKILL.md are vulnerable to command injection. Specifically, the use of unquoted here-documents (e.g., << JSONEOF) allows the shell to perform command substitution on the contents. If the EDIT_PROMPT variable contains malicious shell syntax such as $(command) or backticks, the command will be executed by the host shell during the JSON payload construction.
[EXTERNAL_DOWNLOADS]: The skill uses curl to interact with the official Google Generative Language API (generativelanguage.googleapis.com). This is a well-known service and the usage is consistent with the skill's primary purpose of image generation.
[CREDENTIALS_UNSAFE]: The skill's setup instructions direct users to export their GEMINI_API_KEY into shell profile files (~/.zshrc or ~/.bashrc). This is an insecure practice as these files are stored in plaintext and can be read by other local processes or users, potentially exposing the API credentials.

image-generator