vibe-img
1
总安装量
1
周安装量
#76002
全站排名
安装命令
npx skills add https://github.com/devhashfortheweb/vibe-img --skill vibe-img
Agent 安装分布
windsurf
1
amp
1
cline
1
opencode
1
cursor
1
kimi-cli
1
Skill 文档
vibe-img
AI image generation via <vibe-img> web component. One tag, any provider, cached by default on CDN servers.
When to use this skill
- User asks to build or prototype a website, webapp, or UI component
- User wants visually consistent image sets (cards, features, galleries)
- User wants image editing (img2img, background removal, upscaling, vectorization)
The goal is to speed up the prototyping phase with custom-tailored, non-stock assets.
Setup
Add the script tag in <head>:
<script src="https://cdn.jsdelivr.net/npm/vibe-img@1/dist/vibeimg.js"></script>
No build step. No npm install. Works in any HTML page.
Critical rules
- Always use closing tags.
<vibe-img />breaks the DOM â siblings get swallowed. Always:<vibe-img ...></vibe-img> - Ask the user which provider before generating tags. They need their own API key (
openaiorrecraft). <vibe-img>attributes are cache keys. Every attribute is hashed into a cache key. Changing any attribute â even a single character inpromptâ regenerates the image and discards the cached version. If the user asks for CSS/layout changes only, do not modify<vibe-img>attributes.- Style goes in
img-style, not inprompt. Keep prompts focused on content. - Use
<vibe-theme>for style consistency across 2+ images. It appends a shared style prompt to every child<vibe-img>‘s prompt at generation time. Use it to wrap card grids, feature sections, galleries â anywhere images should look cohesive.
Minimal example
<vibe-img model="recraft" prompt="mountain lake at dawn" img-style="illustration" aspect="wide"></vibe-img>
Provider suggestions
Both providers work for most tasks. If the user doesn’t have a preference:
| Need | Suggestion |
|---|---|
| Illustrations, icons, brand imagery, style control | model="recraft" |
| Photorealistic, complex scenes | model="openai" or model="recraft" with img-style="realistic" |
| Transparent background | recraft + op="remove-bg" or openai + params='{"background":"transparent"}' format="png" |
| SVG output | recraft + op="vectorize" |
| Upscaling | recraft + op="upscale" |
Prompt writing
- Subject first (what is in the image)
- Context (where, when, lighting)
- Composition (angle, framing, focal point)
- Style via
img-styleattribute, NOT in the prompt
Themed groups
<vibe-theme prompt="isometric diorama, clay style, soft studio lighting, pastel palette">
<vibe-img model="recraft" prompt="small coffee shop" aspect="square" img-style="3d"></vibe-img>
<vibe-img model="recraft" prompt="city park with fountain" aspect="square" img-style="3d"></vibe-img>
</vibe-theme>
Image chaining
<vibe-img id="photo" model="recraft" prompt="a cat on a windowsill"></vibe-img>
<vibe-img model="recraft" op="upscale" img-ref="#photo"></vibe-img>
The second element automatically waits for the first to finish before starting. On subsequent page loads both are served from cache instantly â no regeneration, no waiting.
Full attribute and operation reference
See references/api.md for the complete attribute table, all operations, styles, provider-specific params, and common mistakes.