SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsimage-skill-builder

image-skill-builder

Design & creation

View repository

Updated 10-16 08:40

About this skill

Batch Image Generation Skill Creation Assistant

Through structured Q&A, helps users with no coding background create their own custom batch image generation Skill.

Core Concept

Users only need to answer questions; the assistant automatically completes:

  1. Understand user needs → 2. Assemble style templates → 3. Generate complete Skill files → 4. Verify usability

Workflow

Phase 1: Scenario Understanding (3 questions)

Ask questions one by onedo not throw out all questions at once. After each question, wait for the answer.

Q1 - Illustration Scenario:

你想创建什么类型的批量生图 Skill？
  1. 📝 Article/Blog Illustrations — Automatically illustrate each paragraph of a long article
  2. 🎨 Logo/Brand Design — Batch generate Logo variants
  3. 🎬 Storyboard/Storyboard — Break a story into a continuous sequence of frames
  4. 📱 Social Media Images — Xiaohongshu/WeChat Official Account/X illustrations
  5. 🖼️ Poster/Cover Design — Event posters, course covers, Banners
  6. 📊 Data Visualization Images — Infographics, flowcharts, architecture diagrams
  7. 🎭 Character/Character Design — IP mascots, avatars, character design images
  8. ✏️ Other (please describe)

Choose a number, or directly describe your scenario.


**Q2 - 使用场景**:

Where will these images mainly be used?

  1. 📖 Illustrations in articles/documents (need automatic Markdown insertion)
  2. 📲 Social platform publishing (need specific dimensions)
  3. 💼 Brand/commercial use (need high resolution + multiple variants)
  4. 🎓 Teaching/presentation (need to be clear and easy to understand)
  5. 🎮 Entertainment/creation (pursue visual effects)

This affects the image resolution, dimensions, and output format.


**Q3 - 批量数量**:

Each time this Skill is used, about how many images are generated?

  1. 1-3 images (premium route, each one carefully polished)
  2. 5-10 images (standard batch, balancing quality and quantity)
  3. 10+ images (large batch, rapid output)

This affects the Skill's workflow design (whether batch confirmation is needed).


### Phase 2: 风格定义（3-4 个问题）

**Q4 - 视觉风格**:

根据 Q1 的回答，从 references/style-library.md 中筛选出 5-8 个最相关的风格选项展示。

展示格式（每个风格一行描述）：

Based on your scenario, the following styles are recommended:

  1. [Style name] — [One-sentence characteristic] | Suitable for: [Scenario]
  2. [Style name] — [One-sentence characteristic] | Suitable for: [Scenario] ...

Which one do you like? You can also describe the style you want, and I will match it.


**如果用户选了某个风格** → 记录并继续
**如果用户描述了自定义风格** → 提炼关键词，构建自定义风格前缀

**Q5 - 参考图片（可选）**:

Do you have a reference image? Upload a style image you like, and I will break it down:

  • Graphic elements (shape/composition/layout)
  • Color scheme (primary color/secondary color/atmosphere)
  • Visual style (realistic/flat/hand-drawn/3D...)
  • Emotional tone (professional/warm/cool/cute...)
  • Text handling (with or without text/font style)

If not, that's fine too; just say "skip."


**如果用户上传了图片** → 分析图片并生成风格描述，与 Q4 的选择融合
**如果用户跳过** → 继续

**Q6 - 色彩偏好**:

Color preference?

  1. ⬛ Black and white/grayscale — minimalist, professional
  2. 🔵 Monochrome — one primary color + black and white (such as blue tones, green tones)
  3. 🌈 Multicolor — rich color scheme, lively visuals
  4. 🎯 Brand color — Specify a concrete color (give me a color code or description)
  5. 🤷 Whatever — Automatically match based on the style

Pick one, or just tell me the color you want.


**Q7 - 图片比例**:

Image aspect ratio?

  1. 📺 16:9 Landscape — Article images, presentations, banners
  2. ⬜ 1:1 Square — Logos, social avatars, Instagram
  3. 📱 9:16 Portrait — Xiaohongshu, phone wallpapers, Stories
  4. 📄 3:4 Portrait — Posters, covers
  5. 🖥️ 21:9 Ultra-wide — Banners, website headers
  6. Other (tell me the specific ratio)

The most commonly used are 16:9 and 1:1.


### Phase 3: 内容规则（2-3 个问题）

**Q8 - 文字规则**:

Does the image need to include text?

  1. ❌ No text — Purely visual image
  2. 📌 A few keywords — 1-3 core words
  3. 📝 Title + keywords — Clear textual information
  4. 🌐 Mixed Chinese and English — Mainly Chinese, with English for technical terms

If text is needed, use Chinese by default.


**Q9 - 内容来源**:

Where does the image content come from?

  1. ✍️ Describe manually each time — I tell the AI what to draw
  2. 📄 Automatically extract from articles/documents — AI analyzes the article to generate accompanying images
  3. 📋 Batch generate from templates/lists — I provide a list to generate images in bulk
  4. 🖼️ Extend from reference images — Upload an image + description to adapt it

This determines the Skill's input method.


**Q10 - 一致性要求**:

Do multiple images need to maintain a consistent style?

  1. 🔒 Must be consistent — The same visual language (e.g., series illustrations, storyboards)
  2. 🔓 Variation allowed — Style differences are acceptable (e.g., exploring logo variants)
  3. 🎯 Partially consistent — Overall style is consistent but details can vary

The stronger the consistency, the stricter the style anchoring in the prompts.


### Phase 4: 技术配置（快速确认）

**Q11 - 技术参数确认**（一次性展示，用户修改或确认即可）:

Finally, confirm the technical parameters:

Configuration item Current value Description
Resolution 2K Optional 1K/2K/4K
API Gemini (gemini-3-pro-image-preview) Default
Save location obsidian/09image/MMDD-theme-name/ Automatically categorized by date
Insert into document [Yes/No, determined by Q2] Automatically insert into Markdown
Skill name [Automatically generated based on Q1] You can also give it a name

Is there anything you need to modify? If not, just reply "confirm".


### Phase 5: 自动生成 Skill

收集完所有答案后，**自动执行**以下步骤：

#### 5.1 组装 Skill 结构

根据用户回答，决定 Skill 的结构：

| 用户选择 | 生成的 Skill 结构 |
|---------|-----------------|
| 文章配图 | 类似 article-batch-illustration: 分析文章 → 拆段落 → 生成提示词 → 确认 → 批量出图 → 插入文档 |
| Logo 设计 | 类似 logo-batch-generator: 收集品牌信息 → 选风格 → 10个变体 → 出图 → 迭代 |
| 分镜故事 | 类似 storyboard-generator: 润色故事 → 拆分镜 → 选风格 → 批量出图 |
| 社交媒体图 | 新模式: 输入主题/文案 → 选模板 → 批量出图 → 导出特定尺寸 |
| 海报/封面 | 新模式: 输入主题 → 选风格 → 生成多个变体 → 迭代 |
| 数据可视化 | 类似 article-batch-illustration 但专注图表: 输入数据/概念 → 选图表类型 → 出图 |
| 人物/角色 | 新模式: 描述人设 → 选风格 → 多角度/表情变体 → 出图 |

#### 5.2 生成文件

为用户创建完整的 Skill 目录：

skills/[skill-name]/ ├── SKILL.md ← Automatically filled in based on the answers ├── agents/ │ └── openai.yaml ← Automatically generate UI metadata ├── scripts/ │ └── generate_image.py ← Copied from references/generate_image.py └── references/ └── style-guide.md ← Automatically assembled based on the style selection


#### 5.3 生成 SKILL.md

SKILL.md 模板结构（读取 references/skill-template.md 获取完整模板）：

```markdown
---
name: [skill-name]
description: [根据场景+风格自动生成描述，包含触发词]
---

# [Skill 显示名]

[一句话描述功能]

## 工作流程
[根据场景类型填充对应的工作流]

## 风格配置
[用户选定的风格前缀和参数]

## API 配置
[固定 API 信息]

## 保存路径规范
[路径模板]

## 常见用法
[触发词和使用示例]

5.4 Generate style-guide.md

Based on the user's selected style + color + ratio + text rules, extract the complete prompt template for the corresponding style from references/style-library.md, and assemble it into the style-guide.md exclusive to this Skill.

5.5 Copy scripts

Copy scripts/generate_image.py to the new Skill's scripts/ directory.

5.6 Verify

Run quick_validate.py to confirm the Skill structure is correct.

Phase 6: Display and trial

After generation is complete, show the user:

你的 Skill 已创建完成！

📁 位置: skills/[skill-name]/
📄 文件:
  - SKILL.md (主文件)
  - scripts/generate_image.py (图片生成脚本)
  - references/style-guide.md (风格指南)
  - agents/openai.yaml (UI 配置)

🔧 核心配置:
  - 场景: [xxx]
  - 风格: [xxx]
  - 比例: [xxx]
  - 分辨率: [xxx]

想现在试用一下这个 Skill 吗？给我发一段内容，我用刚创建的 Skill 帮你生成图片。

Key Design Principles

  1. Ask only one question at a time — Do not throw multiple questions at once; reduce the user's cognitive load
  2. Provide clear options + custom entry — Numeric selection is fast, while also allowing free-form description
  3. Smart defaults — Infer subsequent defaults based on previous answers (e.g., selecting article illustration → default 16:9, automatically insert into document)
  4. Instant feedback — Briefly confirm understanding after each answer, giving the user a sense of control
  5. Allow skipping and going back — The user can say "skip" or "go back and change item X"

Reference Files

File Purpose When to Read
references/style-library.md Complete prompt template library for all available styles Phase 2, when displaying style options
references/skill-template.md Skill file generation template Phase 5, when assembling the Skill
scripts/generate_image.py Image generation script (copy to new Skill) Phase 5, when copying the script

Common Usages

User Says Action
"Help me create an image generation Skill" Start the complete Q&A flow from Q1
"I want to make a Skill for Xiaohongshu illustrations" Skip Q1 (known scenario = social media), start from Q2
"Create a Skill similar to article illustration, but in cyberpunk style" Skip Q1+Q4 (known scenario + style), quickly confirm other parameters
"Help me change the style of the Skill I just created" Read the existing Skill and only re-run Phase 2