Yummy Gen Image
/install yummy-gen-image
Generate Image
Create or edit images with yummycli gemini nanobanana.
When to Use
Load this skill when the user asks to generate, create, or edit an image using AI — including text-to-image generation, style transfer, image editing with a reference photo, or multi-image compositing.
Prerequisite: Apply the
yummy-sharedskill first.
This skill uses one command for all Gemini image flows:
- Prompt-only generation
- Single-image editing
- Multi-image reference editing
Command Contract
Two equivalent entry points are available:
| Entry point | When to use |
|---|---|
yummycli gemini nanobanana |
Default — human-friendly, Gemini presets applied |
yummycli image generate --provider gemini |
Scripting / automation — explicit, provider-agnostic form |
Both share the same flags and the same Gemini defaults. Prefer gemini nanobanana unless the task explicitly requires the provider-agnostic form.
Basic usage:
yummycli gemini nanobanana --prompt "\x3Cprompt>"
Add one or more reference images when needed:
yummycli gemini nanobanana \
--prompt "\x3Cprompt>" \
--input-image ./source-a.png \
--input-image ./source-b.png
Optional output controls:
--output \x3Cfile>
--model \x3Cmodel>
--aspect-ratio \x3Cratio>
--image-size \x3Csize>
Default values when omitted: --aspect-ratio 16:9, --image-size 1K, --model gemini-3.1-flash-image-preview.
Execution Rules
- Use prompt-only generation when no reference images are provided.
- Use one
--input-imageflag per local image file. - Preserve the user-provided order of reference images.
- Prefer the default model unless the user asks for a specific model.
Model Selection
Use the following model mapping rules:
- If the user explicitly says
gemini pro,pro model, or simplyproin the model-selection context, use:
--model gemini-3-pro-image-preview
- If the user explicitly says
gemini flash,flash model, or simplyflashin the model-selection context, use:
--model gemini-3.1-flash-image-preview
- If the user does not explicitly request a model, omit
--modeland letyummycliuse its default model.
Do not switch models implicitly from vague quality words alone. Only apply the pro or flash mapping when the user's wording clearly refers to model choice.
Model Compatibility
Aspect ratio
| Model | Supported values |
|---|---|
gemini-3.1-flash-image-preview |
1:1 1:4 1:8 2:3 3:2 3:4 4:1 4:3 4:5 5:4 8:1 9:16 16:9 21:9 |
gemini-3-pro-image-preview |
1:1 2:3 3:2 3:4 4:3 4:5 5:4 9:16 16:9 21:9 |
| (default / other) | 1:1 3:4 4:3 9:16 16:9 |
1:4, 1:8, 4:1, 8:1 are flash-only. Do not use them with the pro model.
Image size
| Model | Supported values |
|---|---|
gemini-3.1-flash-image-preview |
512 0.5K 1K 2K 4K |
gemini-3-pro-image-preview |
1K 2K 4K |
| (default / other) | 1K 2K 4K |
512 and 0.5K are flash-only. Do not use them with the pro model.
Intent to Parameters
Translate clear user intent into CLI flags when the mapping is obvious.
Aspect-ratio guidance:
- Use
--aspect-ratio 9:16for requests such as phone wallpaper, vertical poster, story format, or other clearly vertical mobile outputs. - Use
--aspect-ratio 16:9for requests such as desktop wallpaper, presentation cover, horizontal banner, or other clearly widescreen outputs. - Use
--aspect-ratio 1:1for square social images, avatars, or other explicitly square outputs. - If the user already provides a specific ratio, pass it through directly.
- If the output shape is unclear, omit
--aspect-ratioand letyummycliuse its default.
Image-size guidance:
- Use
--image-size 4Kwhen the user explicitly asks for 4K or a clearly print-grade / high-resolution deliverable. - Use
--image-size 2Kwhen the user explicitly asks for 2K or a medium-resolution deliverable. - Use
--image-size 1Kwhen the user explicitly asks for 1K or a lightweight preview-sized result. - Use
--image-size 512or--image-size 0.5Konly when the user explicitly asks for a minimal-size output and the flash model is in use. - If the user does not explicitly request output size, omit
--image-sizeand letyummycliuse its default (1K).
Do not guess --image-size from general quality adjectives alone.
Output path guidance:
- If
--outputis omitted,yummycligenerates a default filename in the current working directory. Do not invent your own output filename unless the user explicitly provides one.
Examples
Prompt-only generation:
yummycli gemini nanobanana \
--prompt "A single ripe banana on a white plate, studio lighting, realistic photo"
Single-image edit:
yummycli gemini nanobanana \
--prompt "Turn this into a watercolor illustration" \
--input-image ./source.png
Multi-image reference edit:
yummycli gemini nanobanana \
--prompt "Blend these references into one polished poster illustration" \
--input-image ./subject.png \
--input-image ./background.jpg
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install yummy-gen-image - 安装完成后,直接呼叫该 Skill 的名称或使用
/yummy-gen-image触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
Yummy Gen Image 是什么?
Use when the user wants to generate or edit raster images with Gemini through yummycli, including prompt-only generation, single-image editing, and multi-ima... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 143 次。
如何安装 Yummy Gen Image?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install yummy-gen-image」即可一键安装,无需额外配置。
Yummy Gen Image 是免费的吗?
是的,Yummy Gen Image 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Yummy Gen Image 支持哪些平台?
Yummy Gen Image 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 Yummy Gen Image?
由 yummysource(@yummysource)开发并维护,当前版本 v1.1.0。