The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input that could influence the behavior of the image generation model.
Ingestion points: The --prompt and --input-image parameters in scripts/generate_image.py accept external content.
Boundary markers: There are no explicit delimiters or instructions provided to the underlying model to ignore embedded instructions in the prompt or image metadata.
Capability inventory: The skill possesses network capabilities (to API providers) and file system access (reading input images and writing output images).
Sanitization: Input prompts and image data are passed to external APIs without sanitization or validation.
[EXTERNAL_DOWNLOADS]: The skill interacts with external services to perform its function.
Well-known services: Communicates with openrouter.ai and Google's GenAI API infrastructure. These interactions are documented and necessary for the skill's primary purpose.

nano-banana-pro