The Agent Skills Directory

[SAFE]: The skill utilizes the official google-genai library and the well-known pillow library for image processing. No suspicious dependencies or external downloads from untrusted sources were found.
[SAFE]: API key handling is implemented using standard environment variables or command-line arguments, with no signs of hardcoded credentials or data exfiltration.
[INDIRECT_PROMPT_INJECTION]: The script accepts arbitrary text prompts and images as input, which are directly used in calls to the Gemini model. This presents an inherent attack surface for indirect prompt injection where malicious input could influence the model's generation logic.
Ingestion points: --prompt and --input-image arguments in scripts/generate_image.py.
Boundary markers: None provided to delimit user input within the model contents.
Capability inventory: Filesystem writes (image output) and network requests (Google AI API).
Sanitization: No sanitization is performed on the user-provided prompt string or image data.

nano-banana-pro