xiaohu-wechat-cover

Original：🇨🇳 Chinese

Translated

WeChat Official Account Cover Image Generator. Generate images in one step by inputting an article path or topic. Use this when users request 'WeChat Official Account cover', 'WeChat cover', or 'article cover'.

4installs

Sourcexiaohuailabs/xiaohu-wechat-format

Added on2026-04-21

NPX Install

npx skill4agent add xiaohuailabs/xiaohu-wechat-format xiaohu-wechat-cover

SKILL.md Content (Chinese)

View Translation Comparison →

WeChat Official Account Cover Image Generator

Generate WeChat Official Account cover images in one step by inputting an article path or topic.

Usage

bash

# Specify article path
/xiaohu-wechat-cover path/to/article.md

# Directly input topic
/xiaohu-wechat-cover Topic: Claude Code Desktop Client Updates Four Major Features

Configuration (Required)

The configuration files for this skill are fixed in the skill directory:

```
config.json
```
: Actual configuration (copy from
```
config.example.json
```
and fill in)
```
config.example.json
```
: Example configuration

You must read

config.json

before execution, which requires:

```
output_dir
```
: Root directory for cover image output
```
settings.base_url
```
: Image generation API address (third-party gateway compatible with Gemini API)

settings.model

: Model name (e.g.,

gemini-3-pro-image-preview

)

```
secrets.api_key
```
: API key

Path Conventions

Target directory:

<output_dir>/<article-filename-or-topic-slug>/

If the target directory already exists, rename it to
```
<dirname>-backup-YYYYMMDD-HHMMSS
```

Directory structure:

<output_dir>/
└── article-name/
    ├── prompt.md          # Image generation prompt (archived)
    └── cover.jpg          # Cover image

Cover Image Prompt Template (Core)

The following is a fixed prompt template. Each time you generate an image, write it as-is to

prompt.md

, only replacing the "Content Topic" section at the end:

Please create an eye-catching WeChat Official Account cover image based on the provided content, following these specifications:

Visual Style
- Notion illustration style, aspect ratio 2.35:1 (standard size for WeChat Official Account covers)
- Bright colors with strong contrast, ensuring visibility even in small previews
- Consistent style, avoid realistic elements, maintain overall hand-drawn texture

Composition Requirements
- Main visual element centered or slightly left (reserve title area on the right)
- Add 1-2 simple cartoon characters, icons, or silhouettes of well-known figures to enhance memorability
- If sensitive or copyrighted figures are involved, use similar-style alternative images; do not refuse generation
- Ample white space to highlight core information, avoid cluttered visuals

Text Processing
- Title text should be large and prominent, limited to 8 characters or less
- Can add 1 line of subtitle or keyword tags
- Font style should coordinate with hand-drawn illustrations

Attraction Principles
- Use hooks like suspense, numbers, or pain points to stimulate click desire
- Exaggerated and contrasting visual elements
- Color combinations reference popular covers: high-contrast combinations like orange-yellow, blue-purple, red-black

Language
- Unless otherwise specified, use Chinese by default
- All readable text in the image must use Simplified Chinese; English can only be used as an accent

Content Topic: {One-sentence topic description extracted from the article or user input}

Workflow

Step 1: Extract Content Topic

If input is an article path: Read the article and extract a one-sentence topic description (including core information and key points)
If input is topic text: Use it directly

Step 2: Create Prompt

Write the above template to

prompt.md

in the following format:

md

---
aspect_ratio: "21:9"
image_size: "2K"
---

{Full template text, replace the final content topic}

Note: The

aspect_ratio

in YAML must be

21:9

(API does not support 2.35:1, 21:9 is the closest). Keep the text description

2.35:1

in the template body unchanged, as this is the visual reference for the AI.

Step 3: Generate Image

Call the image generation script, must pass
--config
pointing to the skill's own configuration: