From m2ai-skills-pack
Activate this skill when the user asks to:
How this skill is triggered — by the user, by Claude, or both
Slash command
/m2ai-skills-pack:banana-makerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Activate this skill when the user asks to:
__pycache__/generate_image.cpython-312.pycgenerate_image.pyoutput/generated_20260318_161715.jpgoutput/test_flash31.jpgoutput/test_flash31_thinking.jpgprompting-guide.mdrequirements.txtvenv/bin/Activate.ps1venv/bin/activatevenv/bin/activate.cshvenv/bin/activate.fishvenv/bin/distrovenv/bin/dotenvvenv/bin/httpxvenv/bin/normalizervenv/bin/pipvenv/bin/pip3venv/bin/pip3.12venv/bin/pyrsa-decryptvenv/bin/pyrsa-encryptActivate this skill when the user asks to:
Take the user's description or prompt as input. It can be a quick idea or a detailed description.
Present these options:
gemini-3.1-flash-image-preview). Fast, cost-effective. Supports 512 to 4K, controllable thinking, new ultra-wide aspect ratios. Good for iteration and quick concepts.gemini-3-pro-image-preview). Advanced reasoning ("Thinking"), high-fidelity text rendering. Best for final output and detailed work. 1K-4K.Read the prompting guide at ~/.claude/skills/banana-maker/prompting-guide.md and apply the Nano Banana Pro methodology to enhance the user's idea into a detailed, high-quality prompt. Claude does this enhancement directly -- no extra API call needed.
Apply the Ultimate Prompt Structure:
Key principles:
Show the enhanced prompt to the user before generating.
Run the script using the banana-maker venv:
~/.claude/skills/banana-maker/venv/bin/python ~/.claude/skills/banana-maker/generate_image.py "THE ENHANCED PROMPT" --model MODEL_CHOICE [OPTIONS]
http://10.0.0.46:PORT/filename if a file server is running, otherwise just the file pathgenerate_image.py PROMPT --model MODEL [OPTIONS]
Positional:
prompt The image generation prompt
Required:
--model, -m flash | pro | grounded
Options:
--output, -o PATH Output file path (default: output/generated_TIMESTAMP.png)
--aspect-ratio, -a AR 1:1 | 16:9 | 9:16 | 4:3 | 3:4 | 1:4 | 4:1 | 1:8 | 8:1 (default: 1:1)
--size, -s SIZE 512 | 1K | 2K | 4K (default: 2K, 512 is Flash 3.1 only)
--reference, -r PATH Reference image path(s), can specify multiple times (max 14)
--search SEARCH web | image | both (requires --model grounded)
--thinking, -t LEVEL minimal | high (Flash 3.1 thinking control, flash/grounded only)
| Flag | Model ID | Notes |
|---|---|---|
flash | gemini-3.1-flash-image-preview | Nano Banana 2: fast, 512-4K, thinking control |
pro | gemini-3-pro-image-preview | Nano Banana Pro: advanced reasoning, 1K-4K |
grounded | gemini-3.1-flash-image-preview | Flash 3.1 + search tools (Nano Banana 2) |
For identity locking, style transfer, or image editing, pass reference images:
~/.claude/skills/banana-maker/venv/bin/python ~/.claude/skills/banana-maker/generate_image.py \
"Transform this photo into a watercolor painting" \
--model flash \
--reference photo.jpg
Supports up to 14 reference images (10 objects + 4 characters for Flash 3.1).
Simple generation:
~/.claude/skills/banana-maker/venv/bin/python ~/.claude/skills/banana-maker/generate_image.py \
"A hyper-realistic close-up of an astronaut fixing a circuit board on Mars. Harsh sunlight from the right creates deep contrast. Shot on 35mm film with visible grain. For a sci-fi documentary poster." \
--model pro --aspect-ratio 16:9
With reference image:
~/.claude/skills/banana-maker/venv/bin/python ~/.claude/skills/banana-maker/generate_image.py \
"Keep this person's facial features identical but place them in a cyberpunk city at night" \
--model pro --reference portrait.jpg
With search grounding (real-world accuracy):
~/.claude/skills/banana-maker/venv/bin/python ~/.claude/skills/banana-maker/generate_image.py \
"The Sydney Opera House at sunset with a dramatic storm approaching from the west" \
--model grounded --search both --aspect-ratio 16:9
Fast low-res iteration with thinking:
~/.claude/skills/banana-maker/venv/bin/python ~/.claude/skills/banana-maker/generate_image.py \
"A minimalist logo for a coffee brand called 'Dark Roast Labs'" \
--model flash --size 512 --thinking high
Ultra-wide panoramic:
~/.claude/skills/banana-maker/venv/bin/python ~/.claude/skills/banana-maker/generate_image.py \
"A sweeping desert landscape at golden hour, endless sand dunes under dramatic cloud formations" \
--model flash --aspect-ratio 8:1 --size 4K
npx claudepluginhub m2ai-mcp-servers/claude-skills --plugin m2ai-skills-packGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.