Version: 1.0 Architecture: Master-Subordinate Learning Mode: Semi-automatic (generate report + manual review)
🎯 Core Features
Automatically extracts reusable elements from prompts in any domain and stores them in the Universal Elements Library database.
Supports 7 major domains:
- 📷 portrait - Portrait Photography
- 🏠 interior - Interior Design
- 📦 product - Product Photography
- 🎨 design - Graphic Design
- 🎭 art - Art Style
- 🎬 video - Video Generation
- 📸 common - General Photography Techniques
📋 Usage
Method 1: Learn a single Prompt
学习这个Prompt: [粘贴完整Prompt]
or
分析并学习: A premium collector's edition book photographed with Phase One camera, featuring Italian calfskin binding...
Method 2: Batch learn 18 source Prompts
批量学习所有Prompts
or
学习extracted_modules.json中的所有Prompts
Method 3: Learn a specific domain
只学习product领域的Prompts
or
学习Prompt #1, #6, #14, #16（产品摄影）
Method 4: Learn design systems/module systems ⭐ New
学习这个设计系统：[粘贴完整的设计系统规范]
or
学习这个工作流模块：[完整工作流内容]
Automatically identifies features：
- Contains keywords:module / system / workflow / framework
- Has a hierarchical structure (background system + color scheme + font system...)
- Contains design philosophy explanations or process steps
Special handling (module system mode)：
- Extract elements → elements table
- Create design system record → design_templates table
- Save complete original content → visual_reference field ⭐ Key enhancement
🔄 Workflow
输入Prompt
↓
【Step 0: 类型识别】⭐ 新增
- 判断是"画面描述型" 还是 "设计系统型"
- 设计系统特征：包含"设计系统"、"风格规范"、有层级结构
- 输出: {"type": "prompt" | "design_system"}
↓
【Step 0.1: 模块系统判断】⭐ 关键增强
- 检测关键词：【模块】【系统】【工作流】【框架】
- YES → 模块系统模式（需保存完整原始内容）
- NO → 普通学习模式（只提取元素）
↓
【Step 1: 领域分类】domain_classifier.md
- 识别主要领域（product/design/art/video...）
- 判断是否多领域
- 输出: {"primary": "product", "secondary": ["photography"]}
↓
【Step 2: 元素提取】element_extractor.md
- 根据领域提取对应元素
- product → product_types, materials, photography_techniques
- design → layouts, effects, typography
- art → art_styles, special_effects
- 输出: List[{category, name, template, keywords}]
↓
【Step 3: 自动打标签】tagger.md
- 基于keywords生成tags
- 添加领域标签、类别标签
- 跨领域标签识别（luxury, glass, dynamic...）
- 输出: tags列表
↓
【Step 4: 计算复用性】
- reusability_score (1-10)
- 基于：通用性、清晰度、独立性
↓
【Step 5: 更新数据库】library_updater.md
- 检查是否已存在（去重）
- 生成element_id
- 写入elements.db
- 更新tag_index
↓
【Step 5.1: 如果是设计系统】⭐ 新增
- 同时创建模板记录
- 写入design_templates表
- 关联所有提取的元素
- 保存设计理念、使用指南等完整信息
↓
【Step 5.2: 如果是模块系统模式】⭐ 关键增强
- 把完整原始内容保存到 visual_reference 字段
- 包括：完整工作流、所有模板、配置参数等
- 目的：可直接使用，无需重新组合元素
↓
【Step 6: 生成报告】
- 新学习的元素列表
- 按领域和类别分组
- 推荐的tags
- 质量评估
- （设计系统）额外显示模板信息
📊 Output Example
Learning Report
# Universal Learner - 学习报告
**学习时间**: 2026-01-01 18:30:00
**源Prompt**: Prompt #1
## 🎯 领域识别
主领域: **product** (产品摄影)
次领域: **common** (通用摄影)
## 📦 提取的元素
### 产品类型 (product_types)
1. **collector_edition_book** - 收藏版书籍
- 模板: premium collector's edition book, luxury binding, Italian calfskin cover
- 关键词: collector's edition, premium book, luxury binding
- 标签: product, book, luxury, collectible
- 复用性: 7.5/10
- element_id: product_product_types_001
### 材质纹理 (material_textures)
2. **glossy_reflective** - 光泽反射材质
- 模板: high-end glossy surface, reflective finish, metallic sheen
- 关键词: glossy, reflective, shiny, polished
- 标签: material, glossy, luxury, product
- 复用性: 8.5/10
- element_id: product_material_textures_002
### 摄影技术 (photography_techniques)
3. **macro_product_shot** - 产品微距特写
- 模板: Phase One medium format camera with 100mm macro lens, razor-sharp focus
- 关键词: macro, close-up, detailed, high-resolution
- 标签: photography, macro, product, technical
- 复用性: 9.0/10
- element_id: common_photography_techniques_032
## ✅ 已添加到数据库
- 3个新元素已写入 elements.db
- 更新了12个标签索引
- product领域: 60 → 63 个元素
- common领域: 31 → 32 个元素
## 💡 质量评估
- 提取完整度: 95%
- 标签质量: 优秀
- 复用性评分: 8.3/10 (平均)
🎛️ Configuration Options
Degree of Automation
- Current: Semi-automatic (generate report + manual review)
- Future: Optional fully automatic mode
Reusability Scoring Criteria
| Score | Criteria |
|---|---|
| 9-10 | Extremely high reusability: cross-domain universal, such as "macro photography", "soft lighting" |
| 7-8 | High reusability: universal within the domain, such as "product on table", "geometric layout" |
| 5-6 | Medium reusability: specific scenarios, such as "collector's edition book" |
| 3-4 | Low reusability: very specific, such as "vintage 1960s typewriter" |
| 1-2 | Extremely low reusability: one-off description, extraction not recommended |
📚 Data Sources
Input: extracted_results/extracted_modules.json (18 source Prompts)
Output:
extracted_results/elements.db(SQLite database)extracted_results/universal_elements_library.json(JSON export)
🔧 Module Description
| Module | File | Function |
|---|---|---|
| Domain Classifier | modules/domain_classifier.md |
Identifies which domain a Prompt belongs to |
| Element Extractor | modules/element_extractor.md |
Extracts reusable elements |
| Tag Generator | modules/tagger.md |
Automatically generates tags |
| Library Updater | modules/library_updater.md |
Updates the database |
✅ Acceptance Criteria
Signs of successful learning:
- ✅ Can correctly identify the 7 major domains
- ✅ Extract ~440 elements from 18 Prompts
- ✅ Automatic deduplication (does not repeatedly add existing elements)
- ✅ High tag quality (strong relevance)
- ✅ Reusability score is reasonable
Skill status: ✅ Implemented Last updated: 2026-01-05 Maintainer: Universal Library System