home / skills / cyangzhou / -2--project-yunshu- / image_generation_expert

image_generation_expert skill

safe

This skill generates high-quality images from text prompts using configurable sizes such as 1024x1024, 768x1344, and 1344x768.

npx playbooks add skill cyangzhou/-2--project-yunshu- --skill image_generation_expert

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

429 B

---
name: Image_Generation_Expert
description: 实现 AI 图像生成能力，支持根据文本描述创建高质量图像。
---

# 图像生成技能

根据文本提示词创建视觉内容。

## 支持尺寸
- 1024x1024 (正方形)
- 768x1344 (人像)
- 1344x768 (风景)

## 代码实现示例
```javascript
const response = await zai.images.generations.create({
  prompt: '阳光下的猫咪',
  size: '1024x1024'
});
```

Overview

This skill provides AI-powered image generation from text prompts, producing high-quality visuals in multiple aspect ratios. It is designed for developers and creators who need fast, programmatic image creation for prototypes, content, or art. Outputs are optimized for common use cases like square, portrait, and landscape formats.

How this skill works

You send a text prompt and optional parameters (size, style hints, seeds) and the skill returns generated image assets or URLs. It supports three preset sizes to match common layouts: 1024x1024 (square), 768x1344 (portrait), and 1344x768 (landscape). The API is asynchronous and returns metadata to manage download, reuse, and attribution.

When to use it

Generate illustration or concept art from a written brief.
Create placeholder or hero images for apps and websites.
Produce portrait-oriented assets for mobile screens or stories.
Generate landscape images for headers, banners, or backgrounds.
Rapidly iterate visual variations for A/B testing or creative review.

Best practices

Write clear, concise prompts with subject, style, and mood (e.g., 'sunlit tabby cat, photorealistic, soft shadows').
Choose the size that matches the final placement to avoid cropping or scaling artifacts.
Use iterative refinement: start broad, then add constraints for color, lighting, or composition.
Respect copyright and avoid prompts that request copyrighted character likenesses or trademarked logos.
Cache generated assets and store prompt metadata so results are reproducible and auditable.

Example use cases

Create a square product mockup (1024x1024) for an e-commerce listing.
Generate a portrait image (768x1344) for a mobile app onboarding screen.
Produce a landscape banner (1344x768) for a website hero section.
Rapidly create visual variations for social media posts and ads.
Prototype concept art from short design briefs during early-stage ideation.

FAQ

What sizes are supported?

Supported presets are 1024x1024 (square), 768x1344 (portrait), and 1344x768 (landscape).

Can I reproduce the same image later?

Yes—store the prompt, parameters, and seed (if provided) to reproduce or refine results.

Is there an example of how to call the API?

A typical call sends a prompt and size parameter and receives image data or URLs in response.