The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/providers/google.ts contains a command injection vulnerability due to unsafe shell interpolation.
Evidence: In the postGoogleJsonViaCurl function, the code executes a curl command using execSync with variables proxyArgs and url embedded directly in the command string: curl -s ... ${proxyArgs} "${url}" ....
Vulnerability: The proxyArgs variable is constructed from proxy-related environment variables (e.g., HTTPS_PROXY), and the url is constructed from GOOGLE_BASE_URL. Neither variable is sanitized or escaped before being passed to the shell. If an attacker can influence these environment variables (for example, by tricking the agent into writing a malicious value to a .env file), they can execute arbitrary shell commands.
[EXTERNAL_DOWNLOADS]: The skill downloads generated images from various AI provider endpoints including OpenAI, Google, DashScope, and the author's own Tuzi API.
Details: The scripts in scripts/providers/ perform HTTP GET requests to retrieve image data from URLs returned by the generation APIs (e.g., api.tu-zi.com, generativelanguage.googleapis.com, api.openai.com, dashscope.aliyuncs.com). This is consistent with the skill's stated purpose of image generation.

tuzi-image-gen