Nano Banana Pro Image Generation & Editing

Problem

When generating images, the friction is not just getting an image, but vague first prompts, drifting edits, and spending too much effort or quota before reaching a usable result. This skill turns image generation and editing into a repeatable script workflow for producing PNG files in a local directory.

How it works

• Generate images: pass the description to --prompt; start with the default 1K output for a quick draft.
• Edit images: pass --input-image and use a prompt that states what to change and what to preserve, such as sky, style, people, or background.
• Iterate: use 1K for fast feedback, then move to 2K or 4K only after composition and subject are locked.
• Naming: outputs follow yyyy-mm-dd-hh-mm-ss-name.png, making each attempt easy to trace.
• Prompt templates: use style, composition, lighting, background, and color palette for vague generation; for edits, explicitly keep subject, crop, lighting, background, text, and overall style.

Notes

• Working directory: run from the user's current directory so the PNG is saved there, not in the skill directory.
• Credentials: use --api-key or GEMINI_API_KEY; missing credentials cause an immediate error.
• Input path: verify that the image path exists and is readable before editing.
• Scope: the skill generates a file and returns its saved path; it does not read image contents or handle quota, permission, or account issues.