# Generate Images With ModelScope Z-Image

Users need a simple way to create images from prompts inside Claude Code or Codex. This skill sends prompts to ModelScope Z-Image-Turbo and saves the generated image locally.

## Install

```bash
npx skillstore add yizhiyanhua-ai/zimage-skill
```

## Metadata

- Status: approved
- Slug: yizhiyanhua-ai-zimage-skill
- Skillstore revision: r1
- Version status: missing
- Tree hash: f70b00cdbe9f1654b80289c039004ae848d0bb81a3033863e9d19b2475b65c50
- Author: yizhiyanhua-ai
- GitHub username: yizhiyanhua-ai
- License: MIT
- Repository: https://github.com/yizhiyanhua-ai/zimage-skill/tree/main/
- Ref: 64ca8af0f54a325752f08bd54e52151061ea659a
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, env\_access, filesystem, external\_commands
- Quality score: 50
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/yizhiyanhua-ai-zimage-skill
- Manifest: https://skillstore.pages.dev/api/skills/yizhiyanhua-ai-zimage-skill/manifest

## Capabilities

- Sends text prompts to the ModelScope Z-Image-Turbo image generation endpoint.
- Polls asynchronous task status for up to 120 seconds.
- Downloads the first returned image and saves it to a local file.
- Uses MODELSCOPE\_API\_KEY from the environment for authentication.
- Supports English and Chinese text prompts.

## Use Cases

- Draft Article Illustrations: Create quick concept images for articles before commissioning final artwork.
- Explore Visual Directions: Generate options for avatars, social posts, app backgrounds, and campaign concepts.
- Test Bilingual Prompts: Compare English and Chinese prompts in one ModelScope workflow.

## Prompt Templates

### Create a Simple Image

```
Generate an image of a calm mountain lake at sunrise and save it as lake.jpg.
```

### Choose a Visual Style

```
Create a watercolor illustration of a small bakery storefront with warm lighting and soft colors.
```

### Specify Composition

```
Generate a square social media image of a robot assistant organizing notes on a clean desk. Leave clear space for a title.
```

### Build a Hero Background

```
Generate a 16:9 hero background for a productivity app. Include a tidy dashboard, soft daylight, diverse devices, and no readable text.
```

## Limitations

- Requires a ModelScope API key and network access.
- Saves only the first returned image from each task.
- Does not edit, upscale, or inpaint existing images.
- Generated results depend on ModelScope service availability and policy checks.

## Best Practices

- Use clear subjects, style, lighting, and composition in each prompt.
- Keep API keys scoped to ModelScope and rotate them if shared by mistake.
- Choose output paths inside the project or a dedicated images folder.

## Anti Patterns

- Do not place real credentials in prompts, screenshots, or shared transcripts.
- Do not use the skill for private or regulated images without provider review.
- Do not save generated files over important local files.

## Security Audit

- Audited at: 2026-07-08T12:24:58.376\+00:00
- Summary: The skill intentionally sends prompts and authenticated API requests to ModelScope, then downloads the returned image. Most static findings are false positives from documentation, Markdown fences, placeholders, a valid PNG icon, and Chinese text entropy. Confirmed risks are external API use, hidden settings storage guidance, environment credential access, and unrestricted output paths.

## Stats

- Views: 220
- Downloads: 59
- Favorites: 0
- Popularity score: 0