Agent Skills
› NeverSight/learn-skills.dev
› gemini-image
gemini-image
GitHub用于根据用户意图生成图像的技能，支持文生图、图生图及多图参考模式。通过读取配置获取密钥，构造提示词后调用API，并返回生成的图片URL。
Trigger Scenarios
用户要求画图或创作图片
用户请求生成艺术作品
用户希望基于文本描述创建图像
Install
npx skills add NeverSight/learn-skills.dev --skill gemini-image -g -y
SKILL.md
Frontmatter
{
"name": "gemini-image",
"description": "Generate images using AI when user wants to create pictures, draw, paint, or generate artwork. Supports text-to-image and image-to-image generation."
}
Gemini Image Generation
Use this skill when user expresses intent to generate images (e.g., "draw a...", "generate an image...", "create a picture...").
Steps
1. Read Configuration
- Read
config/secrets.mdto get API Key
2. Construct Prompt
| Mode | Prompt Format | Example |
|---|---|---|
| Text-to-Image | description text |
a cute orange cat |
| Image-to-Image | image_URL description |
https://xxx.jpg draw similar style |
| Multi-Image Reference | URL1 URL2 description |
https://a.jpg https://b.jpg merge these two |
For image-to-image, upload local images first. See tips/image-upload.md.
3. Call API
curl -s -X POST "https://api.apicore.ai/v1/images/generations" \
-H "Authorization: Bearer API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "model_name",
"prompt": "prompt_text",
"size": "aspect_ratio",
"n": 1
}'
4. Return Result
Extract data[0].url from response and return to user.
Reference Docs
tips/image-upload.md- Image upload methodstips/chinese-text.md- Chinese text handling tips
Version History
- e0220ca Current 2026-07-05 22:59
