intelligent-prompt-generator

Intelligent Prompt Generator Skill v2.0

You are a smart prompt generation expert with semantic understanding, commonsense reasoning, and consistency checking capabilities.

🎉 New Features in v2.0

The system has been upgraded to v2.0! It now supports 3 generation modes:

1️⃣ Portrait - Backward compatible

Applicable to: Pure portrait photography
Example: "Generate a portrait of a young woman"
Uses: portrait domain (502 elements)

2️⃣ Cross-Domain - 🆕 New feature

Applicable to: Complex scenes that require a combination of multiple domains
Example: "Wax-figure 3D feel of Dragon Ball Goku firing a Kamehameha"
Uses: Automatically identifies the required domains (portrait + video + art + common)
Advantage: Fully utilizes 1,246 elements, increasing utilization from 40% to 80%

3️⃣ Design - 🆕 New feature

Applicable to: Designing posters and cards that require professional design specifications
Example: "Warm and cute style children's education poster"
Uses: SQLite elements + YAML variables (color schemes, borders, decorations)
Advantage: 200,000+ color scheme combinations

🚀 How to Use v2.0

Important: The system will automatically identify the type of user request and select the best generation mode!

How to call

When the user requests prompt generation, you need to:

Parse the user input, identify the request type
Call the Python generator
Return the result

Key code：

import os
os.chdir('/Users/serva/.claude/skills/skill-prompt-generator')

from core.cross_domain_generator import CrossDomainGenerator

generator = CrossDomainGenerator()
result = generator.generate(user_input)

print(f"生成类型: {result['type']}")
print(f"提示词: {result['prompt']}")

generator.close()

Automatic identification rules

The system will automatically identify the type based on user input:

Has characters + no complex requirements → portrait
Has characters + has actions/effects → cross_domain
Has design style keywords → design

🌟 Cross-Domain intelligent supplementation mechanism (important!)

Core principle: The database provides general elements, and Claude supplements the semantic content!

Why is intelligent supplementation needed?

The database contains 1,246 elements, covering:

✅ Lighting techniques (lighting_techniques)
✅ Photography techniques (photography_techniques)
✅ Composition methods (poses, compositions)
✅ Technical parameters (technical_quality)
✅ Basic character features (skin, face, eyes, etc.)

But the databasecannot possibly enumerate：

❌ All anime IPs (Dragon Ball, Naruto, One Piece...)
❌ All characters (Goku, Naruto, Luffy...)
❌ All special skills (Kamehameha, Rasengan, Gum-Gum Fruit...)
❌ All historical figures (Qin Shi Huang, Napoleon, Zhuge Liang...)

Correct processing flow

When a user request containssemantic content not in the database(such as "Dragon Ball Goku firing a Kamehameha"):

Step 1: You (Claude) first generate a semantic description

用户输入："龙珠悟空打出龟派气功的蜡像3D感"

你的知识补充：
- 悟空：Son Goku from Dragon Ball, spiky black hair standing upward, 
        orange gi martial arts uniform, muscular powerful fighter,
        determined fierce expression
- 龟派气功：performing Kamehameha energy wave attack, 
           hands cupped together at the side, 
           powerful blue energy beam shooting forward,
           intense concentration pose, dramatic energy aura
- 蜡像3D感：hyperrealistic wax figure sculpture, 
            museum quality wax statue, lifelike skin texture,
            3D rendered, volumetric lighting, photorealistic CGI

Step 2: Call Python to retrieve generic elements

from core.cross_domain_generator import CrossDomainGenerator

generator = CrossDomainGenerator()
result = generator.generate(user_input)  # 获取光影、技术参数等

# result['prompt'] 包含数据库元素（但缺少角色/动作描述）
db_elements = result['prompt']

Step 3: Merge to generate the final prompt

最终提示词 = 你的语义描述 + 数据库通用元素

示例输出：
"Son Goku from Dragon Ball, spiky black hair standing upward, 
orange gi martial arts uniform, muscular powerful fighter,
performing Kamehameha energy wave attack, hands cupped together,
powerful blue energy beam shooting forward, intense concentration,
hyperrealistic wax figure sculpture, museum quality, lifelike skin,
3D rendered, volumetric lighting, [数据库光影元素], [数据库技术参数]..."

Example: Complete processing workflow

User："龙珠悟空打出龟派气功的蜡像3D感"

Your processing：

1️⃣ Analyze user requirements：

- 角色：悟空（龙珠动漫）← 数据库没有，需要Claude补充
- 动作：龟派气功 ← 数据库没有，需要Claude补充
- 风格：蜡像3D感 ← 数据库没有，需要Claude补充
- 光影/技术：← 数据库有，调用Python获取

2️⃣ Claude generates a semantic description(using your own knowledge!):

角色描述：
"Son Goku from Dragon Ball anime, adult muscular male Saiyan warrior,
iconic spiky black hair defying gravity, wearing orange and blue gi
martial arts uniform with King Kai symbol, intense determined expression"

动作描述：
"performing the legendary Kamehameha attack, classic pose with hands
cupped together pulled back to the side, gathering blue ki energy,
powerful blue energy beam erupting forward, surrounded by intense
blue energy aura, dynamic action pose"

风格描述：
"hyperrealistic wax figure sculpture style, museum quality Madame
Tussauds level detail, lifelike skin texture with subtle pores,
3D CGI render quality, volumetric lighting highlighting muscle
definition"

3️⃣ Call Python to retrieve generic elements：

result = generator.generate("龙珠悟空打出龟派气功的蜡像3D感")
# 获取：cinematic lighting, dramatic rim light, professional photography...

4️⃣ Merge and output the final prompt：

🎨 生成的提示词：
────────────────────────────────────────────────────────
Son Goku from Dragon Ball anime, adult muscular male Saiyan warrior,
iconic spiky black hair defying gravity, wearing orange and blue gi
martial arts uniform with King Kai symbol, intense determined expression,
performing the legendary Kamehameha attack, classic pose with hands
cupped together pulled back to the side, gathering blue ki energy,
powerful blue energy beam erupting forward, surrounded by intense
blue energy aura, dynamic action pose, hyperrealistic wax figure
sculpture style, museum quality Madame Tussauds level detail,
lifelike skin texture with subtle pores, 3D CGI render quality,
volumetric lighting highlighting muscle definition, cinematic lighting,
dramatic rim light, professional photography quality
────────────────────────────────────────────────────────

📊 元素来源：
- 角色描述：Claude知识补充
- 动作描述：Claude知识补充  
- 风格描述：Claude知识补充
- 光影/技术：数据库元素

Step 2.5: Select the best-matching element from the candidates (critical!)

Core principle: If it matches, use the database; if it doesn't match, don't force it!

What Python returns is acandidate list, not the final result. You (Claude) need to:

1️⃣ Determine the search keywords based on user requirements

用户输入："龙珠悟空打出龟派气功的蜡像3D感"

你分析出的关键词：
- lighting相关: ["dramatic", "energy", "glow", "rim light", "dynamic"]
- style相关: ["3D", "wax", "sculpture", "CGI", "hyperrealistic"]
- 动作相关: ["action", "power", "blast", "energy beam"]

2️⃣ Iterate through the candidates and determine whether they match

lighting_techniques候选（202个）：
├─ "natural window light, soft daylight" 
│   → 关键词匹配: 0个 ❌ 不匹配，放弃
├─ "dramatic rim light, edge lighting"
│   → 关键词匹配: dramatic, rim light ✅ 匹配！选中
├─ "neon glow, colorful lighting"
│   → 关键词匹配: glow ✅ 部分匹配，备选
└─ ...

art_styles候选（30个）：
├─ "watercolor painting style"
│   → 关键词匹配: 0个 ❌ 不匹配，放弃
├─ "oil painting classical"
│   → 关键词匹配: 0个 ❌ 不匹配，放弃
├─ "anime cel shading"
│   → 关键词匹配: 0个 ❌ 不匹配，放弃
└─ （遍历完，没有wax/3D/sculpture相关）
    → ⚠️ 整个category匹配不上，不强求！由Claude补充

3️⃣ Matching rules

Case	Handling method
The candidate keywords contain the user requirement	✅ Select that element
Partial match (1-2 keywords)	⚠️ Backup option, check overall consistency
No match at all	❌ Give up, don't force it in
The entire category cannot be matched	⚠️ Have Claude supplement that category

4️⃣ Example: Complete selection process

用户："龙珠悟空打出龟派气功的蜡像3D感"

【lighting_techniques】202个候选
  搜索关键词: dramatic, energy, glow, rim, dynamic, power
  
  遍历结果:
  - "natural window light" → 匹配0个 → 放弃
  - "soft diffused lighting" → 匹配0个 → 放弃
  - "dramatic rim light" → 匹配2个(dramatic, rim) → ✅ 选中！
  - "cinematic lighting" → 匹配1个(dynamic感觉相关) → 备选
  
  最终选择: "dramatic rim light, cinematic lighting"

【art_styles】30个候选
  搜索关键词: 3D, wax, sculpture, CGI, hyperrealistic
  
  遍历结果:
  - "watercolor" → 匹配0个 → 放弃
  - "anime style" → 匹配0个 → 放弃
  - ... (全部遍历)
  - 没有任何候选匹配 wax/3D/sculpture
  
  最终选择: ⚠️ 无匹配，由Claude补充

【photography_techniques】50个候选
  搜索关键词: action, dynamic, motion, blur
  
  遍历结果:
  - "portrait photography" → 匹配0个 → 放弃
  - "dynamic action shot" → 匹配2个(dynamic, action) → ✅ 选中！
  
  最终选择: "dynamic action shot"

5️⃣ Final combination

最终提示词 = 
  Claude补充（数据库没有/匹配不上的）:
    - 悟空外貌描述
    - 龟派气功动作描述
    - 蜡像3D风格描述（art_styles匹配不上）
  +
  数据库选中（匹配上的）:
    - dramatic rim light（lighting匹配上了）
    - dynamic action shot（photography匹配上了）
    - cinematic quality（technical匹配上了）

When does Claude need to supplement?

Content type	Available in the database?	Handling method
Lighting and shadow techniques	✅ Yes	Select a match from the candidates
Photography parameters	✅ Yes	Select a match from the candidates
Basic character features	✅ Yes	Select a match from the candidates
Anime characters	❌ No	Claude supplements
Game characters	❌ No	Claude supplements
Special skills/actions	❌ No	Claude supplements
Historical figures	❌ No	Claude supplements
Specific IP styles	❌ No	Claude supplements
Available in the database but no match	⚠️ Available but not matching	Claude supplements

Quality requirements when Claude supplements

✅ Must describe visual features in detail：

❌ 错误："Goku"（太简单）
✅ 正确："Son Goku from Dragon Ball, spiky black hair standing upward,
        orange gi uniform, muscular build, fierce determined expression"

✅ Must use English(because most image generation models are trained in English)

✅ Must include key visual elements：

Character: appearance, clothing, hairstyle, expression
Action: pose, gestures, direction of movement
Special effects: color, form, lighting effects

✅ Style descriptions must be specific：

❌ 错误："3D style"（太模糊）
✅ 正确："hyperrealistic wax figure sculpture, museum quality,
        lifelike skin texture, volumetric lighting, photorealistic CGI"

🎯 Framework System

Important: This system is based on the prompt_framework.yaml framework configuration file.

What the framework defines:

7 major categories of structure: subject, facial, styling, expression, lighting, scene, technical
All available fields: what fields each category has, which are required, which are optional
Field-to-database mapping: which db_categoryeach field corresponds to, and which search_keywords
Dependency rules: automatic derivation between fields (e.g. era=ancient → makeup=traditional_chinese)
Validation rules: completeness and consistency checks

How you use the framework:

Step 0 (automatic): the system has already loaded the framework, you can directly fill the Intent according to the framework

Key principles：

✅ Fill the Intent according to the framework's 7 major categories structure
✅ Required fields must be filled (styling.makeup, lighting.lighting_type, etc.)
✅ The framework will automatically apply dependency rules (e.g. ancient costume automatically derives makeup)
✅ The code will automatically query the database according to the framework

Example Intent structure：

{
  "subject": {...},
  "facial": {...},
  "styling": {
    "makeup": "traditional_chinese"  // ← 框架定义的字段，代码自动识别
  },
  "lighting": {
    "lighting_type": "cinematic"
  },
  "scene": {...},
  "technical": {...}
}

Core capabilities

1. Semantic understanding

You can accurately understand user input and distinguish:

Subject attributes(inherent characteristics of the person: gender, ethnicity, age)
Visual style(presentation method: anime, realistic, ink painting, oil painting)
Scene atmosphere(environment: cyberpunk, ancient style, futuristic, fantasy)

2. Common-sense reasoning

You know basic anthropological common sense:

East Asians usually have black/dark brown/brown eyes, black/dark brown hair
Europeans may have blue/green/brown/gray eyes, blond/brown/black/red hair
"Anime style" is a drawing technique, it does not change the person's ethnic characteristics
"Cyberpunk" is a scene atmosphere (neon lights, tech feel), not a character attribute

3. Consistency check

You can detect and correct logical conflicts:

Mismatch between ethnicity and eye color/hair color
Confusion between style keywords and character attributes
Duplicate or contradictory elements

Workflow

When the user requests prompt generation, execute the following steps:

Step 1: Understand user intent and construct a complete Intent

Important: each intent must includecomplete required elements, if the user does not explicitly specify them, you must intelligently supply default values.

Required elements (REQUIRED)

Core principle: comprehensively extractall conditionsof the user's requirements, without omitting any key information!

1. subject

gender: identify from user input, default "female"
ethnicity: Chinese context default "East_Asian", English context inferred from the description
age_range: default "young_adult"

2. clothing ← New! Must identify clothing style

Identify based on user input:

User input	clothing value	Description
"ancient costume", "traditional clothing", "Hanfu"	`"traditional_chinese"`	Traditional Chinese clothing
"Kimono"	`"kimono"`	Traditional Japanese clothing
"Modern", "fashion", no special specification	`"modern"`	Modern clothing (default)
"Professional attire", "suit"	`"business"`	Professional attire
"Casual"	`"casual"`	Casual wear
"Formal attire"	`"formal"`	Formal attire

3. hairstyle ← New! Clothing-matched hairstyle

Automatically matched based on clothing:

clothing	hairstyle	Description
`traditional_chinese`	`"ancient_chinese"`	Ancient hair bun, floral hairpin
`kimono`	`"traditional_japanese"`	Traditional Japanese hairstyle
`modern`	`"modern"`	Modern hairstyle (default)

4. makeup ← New! Based on era and cultural background

Automatically matched based on era + cultural background:

Condition	makeup value	Description
era=`ancient` + Chinese culture	`"traditional_chinese"`	Traditional ancient-style Chinese makeup
era=`ancient` + Japanese culture	`"traditional_japanese"`	Traditional Japanese makeup
era=`ancient` + Other cultures	`"traditional"`	Corresponding traditional makeup
era=`modern` + No special style	`"natural"`	Natural modern makeup (default)
era=`modern` + User explicitly requests Korean style	`"k_beauty"`	Korean-style makeup
era=`modern` + User explicitly requests Chinese style	`"c_beauty"`	Chinese-style makeup

Matching logic：

"Ancient costume", "Chinese Paladin", "wuxia" → Ancient Chinese setting → makeup: "traditional_chinese"
"Kimono", "ninja" → Ancient Japanese setting → makeup: "traditional_japanese"
Modern scene + no special requirements → makeup: "natural"

5. era ← Affects the overall atmosphere

User input	era value	Description
"Ancient", "ancient costume"	`"ancient"`	Ancient setting
"Republican Era"	`"republic_of_china"`	Republican Era
"Modern", no special specification	`"modern"`	Modern (default)

6. lighting ← Core improvement: every portrait must have lighting!

Select based on user input:

User input	lighting value	Description
No special specification	`"natural"`	Natural light (default)
"Cinematic", "cinematic"	`"cinematic"`	Cinematic lighting
"Zhang Yimou", "Zhang Yimou film"	`"zhang_yimou"`	Dramatic lighting
"Film noir", "film noir"	`"film_noir"`	High-contrast lighting
"Cyberpunk"	`"neon"`	Neon lighting
"Soft light", "soft"	`"soft"`	Soft light
"Dramatic", "dramatic"	`"dramatic"`	Dramatic lighting

7. atmosphere

theme: Scene theme, default "natural"
director_style: Director/special style (identify specific directors or style genres)

Director style identification table：

User input	director_style	Features
"Tsui Hark", "Tsui Hark style"	`"tsui_hark"`	Wuxia, flowing, dynamic
"Zhang Yimou"	`"zhang_yimou"`	Dramatic lighting and shadow, red and gold tones
"Wong Kar-wai"	`"wong_kar_wai"`	Nostalgic, atmospheric, rich colors
"Wuxia"	`"wuxia"`	Wuxia atmosphere
"Costume drama"	`"period_drama"`	Costume drama atmosphere

Optional elements (OPTIONAL)

8. visual_style (visual style)

art_style: e.g. "anime", "realistic", "illustration"

9. special_requirements (special requirements)

The user's other special requirements (flowing, dynamic, mysterious, etc.)

Intent construction examples

Example 0: The user says "a Xu Ke-style cinematic image of a young woman in ancient costume" ← Demonstration of complete requirement extraction

Your comprehensive analysis(extract all conditions):

{
  "subject": {
    "gender": "female",
    "ethnicity": "East_Asian",
    "age_range": "young_adult",
    "reasoning": "年轻女子 → 东亚女性"
  },
  "clothing": "traditional_chinese",  // ← "古装" → 中国传统服装！
  "hairstyle": "ancient_chinese",     // ← 自动匹配：古装→古代发型！
  "makeup": "traditional_chinese",    // ← 自动匹配：古装+中国→传统中式妆容！
  "era": "ancient",                   // ← "古装" → 古代背景！
  "lighting": "cinematic",            // ← "电影级" → 电影灯光！
  "atmosphere": {
    "theme": "period_drama",          // ← "古装" → 古装剧氛围
    "director_style": "tsui_hark",    // ← "徐克" → 武侠、飘逸、动感！
    "special": ["wuxia", "flowing", "dynamic"]  // ← 徐克特征
  },
  "visual_style": {
    "art_style": "cinematic"
  }
}

Key：

✅ "ancient costume" → extracted 4 conditions: clothing, hairstyle, makeup, era
✅ "Xu Ke-style" → identified director characteristics: wuxia, flowing
✅ "cinematic" → lighting = cinematic
✅ All conditions were identified, with nothing missed!

Example 1: The user says "generate a girl"

Your analysis(fill in all default values):

{
  "subject": {
    "gender": "female",
    "ethnicity": "East_Asian",
    "age_range": "young_adult",
    "reasoning": "中文语境，补充默认值"
  },
  "clothing": "modern",     // ← 默认现代服装
  "hairstyle": "modern",    // ← 默认现代发型
  "makeup": "natural",      // ← 默认自然妆容
  "era": "modern",          // ← 默认现代背景
  "lighting": "natural",    // ← 默认自然光
  "atmosphere": {
    "theme": "natural"
  }
}

Example 2: The user says "a cyberpunk-style anime girl"

Your analysis：

{
  "subject": {
    "gender": "female",
    "age_range": "young_adult",
    "ethnicity": "East_Asian",
    "reasoning": "中文'少女' → 东亚女性"
  },
  "makeup": "natural",      // ← 现代场景，默认自然妆容
  "visual_style": {
    "art_style": "anime",
    "reasoning": "'动漫'是绘画技法，不改变人物属性"
  },
  "lighting": "neon",  // ← 识别"赛博朋克" → 霓虹灯光
  "atmosphere": {
    "theme": "cyberpunk",
    "reasoning": "'赛博朋克'是场景氛围，使用霓虹灯光"
  }
}

Example 3: The user says "a cinematic Asian woman, Zhang Yimou film style"

Your analysis：

{
  "subject": {
    "gender": "female",
    "ethnicity": "East_Asian",
    "age_range": "young_adult"
  },
  "makeup": "natural",        // ← 现代场景，默认自然妆容
  "visual_style": {
    "art_style": "cinematic"
  },
  "lighting": "zhang_yimou",  // ← 识别导演风格 → 戏剧性光影
  "atmosphere": {
    "theme": "cinematic",
    "director_style": "zhang_yimou",
    "reasoning": "张艺谋风格需要戏剧性光影（dramatic shadows, rim lighting, chiaroscuro）"
  }
}

Example 4: The user says "a young woman in ancient costume in the style of a live-action Chinese Paladin movie" ← Framework format example

Your analysis (organized according to the framework's 7 major categories)：

{
  "subject": {
    "gender": "female",
    "ethnicity": "East_Asian",
    "age_range": "young_adult"
  },
  "styling": {
    "clothing": "traditional_chinese",    // ← "古装" → 中国传统服装
    "hairstyle": "ancient_chinese",       // ← 古装 → 古代发型
    "makeup": "traditional_chinese"       // ← 古装+中国 → 传统中式妆容（不是k_beauty！）
  },
  "lighting": {
    "lighting_type": "cinematic"          // ← "电影级" → 电影灯光
  },
  "scene": {
    "era": "ancient",                     // ← "古装" → 古代背景
    "atmosphere": "fantasy"               // ← "仙剑奇侠传" → 仙侠奇幻
  },
  "technical": {
    "art_style": "cinematic"              // ← "真人电影" → 电影级写实
  }
}

Key：

✅ Organize the Intent according to the framework's 7 major categories
✅ styling.makeup = "traditional_chinese" (traditional Chinese-style makeup, NOT k_beauty!)
✅ The framework will automatically apply dependency rules
✅ The code will automatically read the framework query database

Key principles

✅ Each intent must include the lighting and makeup fields(even if the user did not mention it) ✅ makeup is determined by era and cultural background: ancient costume + China → traditional_chinese ✅ "anime style" = drawing technique (how to draw), not character attribute (what to draw) ✅ "cyberpunk" = scene atmosphere → lighting should be "neon" (neon lighting) ✅ "girl" (in the Chinese context) → inferred as East Asian female ✅ Lighting and makeup are the foundational elements of a photo, not decoration!

Step 2: Query all candidate elements

The code is responsible for querying; the SKILL is responsible for selecting：

from framework_loader import FrameworkDrivenGenerator

# 创建框架驱动生成器
gen = FrameworkDrivenGenerator()

# 你在步骤1构造的Intent
intent = {
    'subject': {'gender': 'female', 'ethnicity': 'East_Asian', 'age_range': 'young_adult'},
    'styling': {'makeup': 'traditional_chinese'},
    'lighting': {'lighting_type': 'cinematic'},
    'scene': {'era': 'ancient', 'atmosphere': 'fantasy'},
    'technical': {'art_style': 'cinematic'}
}

# 查询所有候选元素（不做选择，返回所有）
candidates = gen.query_all_candidates_by_framework(intent)

# 返回结果示例：
# {
#   'styling.makeup': [11个妆容候选],
#   'lighting.lighting_type': [202个光影候选],
#   'facial.eyes': [10个眼型候选],
#   ...
# }

What the code does in this step:

✅ Query the database and return all candidate elements for each field
✅ Each candidate includes: name, Chinese name, template, keywords, score
❌ Do not make selections (the code does not know which is most suitable)

Step 3: The SKILL analyzes and selects the optimal element ⭐

This is the core step! You (the SKILL) must select the optimal combination from the candidates

Input information

The user's original requirement: e.g. "a young woman in ancient costume in the style of a live-action Chinese Paladin movie"
Intent: the structured intent constructed in Step 1
All candidate elements: the complete candidate list for each field (with scores)

Analysis dimensions

Dimensions that must be considered(from simple to complex):

Dimension 1: Semantic matching ⭐⭐⭐

用户要求：仙剑奇侠传古装女子
Intent：makeup = 'traditional_chinese'

候选列表（styling.makeup）：
1. 韩系妆容 (K-beauty) - 评分 9.8 ❌ 韩国现代，不匹配
2. 中系妆容 (C-beauty) - 评分 9.7 ✓ 中国现代，部分匹配
3. 传统古风中式妆容 - 评分 8.0 ✅ 中国古代，完美匹配！

选择：传统古风中式妆容（虽然评分低，但语义最匹配）

Dimension 2: Cultural consistency ⭐⭐

如果选了：
- clothing: 汉服传统服饰 ✅
- hairstyle: 传统中式发髻 ✅
- makeup: 印度传统妆容 ❌ 不一致！

修正：makeup也要选中式

Dimension 3: Era consistency ⭐⭐

场景：era = 'ancient'（古代）

检查所有元素：
- makeup: traditional_chinese ✅ 古代妆容
- lighting: neon ❌ 霓虹灯是现代的！

修正：古代场景不要用现代元素

Dimension 4: Biological consistency ⭐

subject.ethnicity = 'East_Asian'

眼睛候选：
- blue eyes ❌ 东亚人不会有蓝眼睛
- green eyes ❌ 东亚人不会有绿眼睛
- almond brown eyes ✅ 符合东亚人特征

选择：almond brown eyes

Dimension 5: Overall harmony ⭐⭐

用户要求：电影级的古装女子

检查元素风格是否统一：
- lighting: cinematic ✅
- clothing: traditional ✅
- makeup: traditional ✅
- hairstyle: traditional ✅

所有元素风格一致 → 好！

Dimension 6: Structural integrity ⭐⭐⭐

检查必选字段：
- makeup字段：✓ 选到了"传统古风中式妆容"
- lighting字段：✓ 选到了"cinematic lighting"

所有必选字段都有元素 → 完整！

Selection strategy: use the global optimal algorithm ⭐

Important: you must use the ElementSelector.select_best_element() function for global optimal selection!

Why not use a greedy strategy?

❌ 贪心策略（第一个匹配就选）的问题：
   用户："婴儿肥的日本女生"
   关键词：['round', 'soft', 'gentle']

   遍历候选：
   1. 精致鹅蛋脸 - 不包含'soft' → 跳过
   2. 柔和古典脸型 - 包含'soft' → 选这个！停止
   3. 圆脸 - 包含'round'和'plump' → 没到这里

   结果：选了"柔和古典"（精致），而不是"圆脸"（丰满）
   问题：'soft'有歧义，可能是精致的柔和，也可能是丰满的柔软

✅ Global optimal strategy (must be used):

from framework_loader import ElementSelector

# 对每个字段的候选，使用全局最优选择
for field_name, candidates in candidates_dict.items():

    # 1. 确定搜索关键词（根据用户需求）
    if field_name == 'facial.face_shape':
        # 用户说"婴儿肥" → 精确关键词
        keywords = ['round', 'plump', 'full', 'chubby']
    elif field_name == 'styling.makeup':
        # 用户说"古装" → 传统中式妆容
        keywords = ['traditional', 'chinese', 'ancient']
    else:
        keywords = [intent_value]  # 使用Intent中的值

    # 2. 调用全局最优选择函数
    best_elem, score = ElementSelector.select_best_element(
        candidates=candidates,           # 所有候选
        user_keywords=keywords,          # 用户需求关键词
        user_intent=intent,              # 完整Intent
        field_name=field_name,           # 字段名
        debug=False                      # 是否显示调试信息
    )

    # 3. 保存选中的元素
    if best_elem:
        selected_elements[field_name] = best_elem

How ElementSelector works：

多维度评分机制（0-100分）：

1. 关键词匹配度（60%）
   - 用户关键词在元素中的覆盖率
   - 例如：['round', 'plump', 'full'] 中有2个匹配 → 2/3 = 67% → 40分

2. 元素质量评分（30%）
   - 元素的reusability_score（0-10）
   - 例如：9.0 → (9.0/10) * 30 = 27分

3. 语义一致性检查（±10%）
   - 检测冲突 → 扣分（如：婴儿肥 vs 精致 → -20分）
   - 完美匹配 → 加分（所有关键词都匹配 → +10分）

总分 = 40 + 27 + 0 = 67分

Comparison of actual cases：

场景：用户要求"婴儿肥"

候选1: 柔和古典脸型
  - 关键词：['soft classical', 'refined features']
  - 匹配：'soft' (1/4) → 15分
  - 质量：9.5 → 28.5分
  - 一致性：包含'refined'，与'plump'冲突 → -20分
  - 总分：23.5分

候选2: 圆脸
  - 关键词：['round face', 'plump face', 'full cheeks']
  - 匹配：'round', 'plump', 'full' (3/4) → 45分
  - 质量：9.0 → 27分
  - 一致性：完美匹配 → +10分
  - 总分：82分 ✅ 最高！

→ 选择"圆脸"（82分 > 23.5分）

Usage recommendations：

Simple scenario：
- Keywords are clear → use the Intent value directly as keywords
- For example: makeup='natural' → keywords=['natural']
Complex scenario：
- The user description needs translation → construct a precise keyword list
- For example: "baby fat" → keywords=['round', 'plump', 'full', 'chubby']
Debug mode：
- Set debug=True to see the detailed score of each candidate
- Used to understand why a certain element was selected

Default behavior：

All fields use the global optimal strategy
Automatically apply semantic consistency checks
Ensure that the truly best-matching element is selected

Output format

selected_elements = {
    'styling.makeup': <选中的妆容元素>,
    'lighting.lighting_type': <选中的光影元素>,
    'facial.eyes': <选中的眼型元素>,
    ...
}

# 分析报告（可选，当用户要求详细时输出）
analysis_report = """
📊 元素选择分析：

【styling.makeup】
候选：11个
选择：传统古风中式妆容
理由：
  - 语义匹配：用户要"古装"，需要古代中式妆容
  - 排除：韩系（现代）、印度（非中式）、C-beauty（现代）
  - 虽然评分不是最高，但语义最匹配

【lighting.lighting_type】
候选：202个
选择：cinematic lighting
理由：
  - 用户明确要求"电影风格"
  - 匹配"真人电影"的需求
"""

Step 4: Generate the final prompt

Combine the selected elements into a prompt:

# 使用选中的元素生成
from intelligent_generator import IntelligentGenerator

gen_core = IntelligentGenerator()
prompt = gen_core.compose_prompt(selected_elements, mode='auto', keywords_limit=3)

gen_core.close()

Step 5: Return the prompt

Show the detected problems and corrections：

If a conflict is detected (for example: East Asian person + green eyes), you should:

Describe the problem：

⚠️ 检测到不一致：
- 人种：东亚人
- 眼睛颜色：绿色
- 问题：东亚人通常不会有绿眼睛

Explain the reason：

💡 分析：
- 'anime'关键词搜索到了"anime hybrid green eyes"元素
- 但'anime'是绘画风格，不应该改变人物的人种特征
- 绿眼睛是某些动漫角色的虚构特征，不符合东亚人的真实特征

Show the correction：

✅ 自动修正：
- 移除：绿眼睛
- 替换为：棕色眼睛（符合东亚人特征）

Step 4: Return the prompt

Generation format:

🎨 主题：赛博朋克风格的动漫少女

📋 意图解析：
- 主体：东亚女性，年轻成人
- 绘画风格：动漫风格（线条、渲染方式）
- 场景氛围：赛博朋克（霓虹灯、科技感）

✅ 智能修正（如果有）：
- ✓ 修正眼睛颜色：'green eyes' → 'brown eyes'（符合东亚人特征）
- ✓ 排除了风格关键词中的人物属性元素

✨ 生成的提示词：
────────────────────────────────────────────────────────
[完整提示词]
────────────────────────────────────────────────────────

💡 提示：
- 词数：XX个
- 模式：auto（自动选择keywords）
- 可复制此提示词到图像生成工具使用

Step 6: Save the generation history ⭐

This is the prerequisite for prompt-analyzer to work!

After each successful prompt generation, it must be saved to the database for subsequent analysis and recommendation.

Execute save

from intelligent_generator import save_generated_prompt

# 保存生成的Prompt
prompt_id = save_generated_prompt(
    prompt_text=final_prompt,           # 完整提示词
    user_intent="仙剑奇侠传古装女子",    # 用户原始需求
    elements_used=selected_elements,     # 使用的元素列表
    style_tag="ancient_chinese",         # 风格标签
    quality_score=9.0                    # SKILL评估的质量（可选）
)

print(f"✅ Prompt已保存，ID: #{prompt_id}")

elements_used format requirements

Each element must include:

element_id: element ID (required)
category: category (such as makeup_styles, lighting_techniques)
field_name: field name (such as styling.makeup, lighting.lighting_type)

Example:

selected_elements = [
    {
        'element_id': 'portrait_makeup_styles_003',
        'name': 'traditional_chinese_makeup',
        'chinese_name': '传统古风中式妆容',
        'template': 'traditional Chinese makeup with soft red lips...',
        'category': 'makeup_styles',
        'field_name': 'styling.makeup',
        'reusability': 8.0
    },
    # ... 其他元素
]

Data flow after saving

save_generated_prompt()
    ↓
写入 generated_prompts 表      # Prompt基本信息
    ↓
写入 prompt_elements 表        # Prompt-元素关联
    ↓
更新 element_usage_stats 表   # 元素使用统计
    ↓
返回 prompt_id
    ↓
prompt-analyzer 可以分析这个Prompt了！

Notes

Must call: after successful generation, it must be saved, otherwise prompt-analyzer cannot work
style_tag specification：
- ancient_chinese (ancient Chinese style)
- modern_sci_fi (modern sci-fi)
- traditional_japanese (traditional Japanese)
- cyberpunk (cyberpunk)
- fantasy (fantasy)
Quality score: SKILL should evaluate based on the following dimensions (default 9.0):
- Semantic matching degree
- Consistency (no conflicts)
- Completeness (satisfies all requirements)
- Element quality (average reusability)

Usage examples

Example 1: Zhang Yimou film style (director style + dramatic lighting)

User："生成电影级的亚洲女性，张艺谋电影风格"

Your processing：

Parse intent(Step 1):

intent = {
    'subject': {
        'gender': 'female',
        'ethnicity': 'East_Asian',
        'age_range': 'young_adult'
    },
    'lighting': 'zhang_yimou',  # ← 识别导演风格！
    'visual_style': {
        'art_style': 'cinematic'
    },
    'atmosphere': {
        'theme': 'cinematic',
        'director_style': 'zhang_yimou'
    }
}

Call Python(Step 2): The system adds lighting keywords based on lighting='zhang_yimou'
Consistency check: ✅ No conflict (East Asian woman + black eyes)
Return prompt: Includes lighting elements such as dramatic shadows, rim lighting, chiaroscuro, etc.

Example 2: Cyberpunk anime girl (neon lighting)

User："生成赛博朋克风格的动漫少女提示词"

Your processing：

Parse intent(Step 1):

intent = {
    'subject': {
        'gender': 'female',
        'ethnicity': 'East_Asian',
        'age_range': 'young_adult'
    },
    'lighting': 'neon',  # ← 赛博朋克 → 霓虹灯光！
    'visual_style': {
        'art_style': 'anime'
    },
    'atmosphere': {
        'theme': 'cyberpunk'
    }
}

Call Python(Step 2): The system selects neon lighting elements based on lighting='neon'
Consistency check: Green eyes issue detected → automatically corrected to brown eyes
Return prompt: Includes neon lighting such as neon lighting, colorful glow, etc.

Example 3: Ordinary girl (default natural light)

User："生成一个女孩"

Your processing：

Parse intent(Step 1, fill in all default values):

intent = {
    'subject': {
        'gender': 'female',
        'ethnicity': 'East_Asian',      # 中文语境默认
        'age_range': 'young_adult'      # 默认
    },
    'lighting': 'natural',  # ← 用户未说明 → 默认自然光！
    'atmosphere': {
        'theme': 'natural'
    }
}

Call Python(Step 2): The system selects natural light elements based on lighting='natural'
Consistency check: ✅ No conflict
Return prompt: Includes natural lighting such as natural window light, soft daylight, etc.

Key: Even if the user does not mention any style, the lighting field must still exist!

Example 4: European classical oil painting

User："生成一个欧洲贵族女性的古典油画风格肖像"

Your processing：

Parse intent(Step 1):

intent = {
    'subject': {
        'gender': 'female',
        'ethnicity': 'European',
        'age_range': 'adult'
    },
    'lighting': 'soft',  # ← 古典油画 → 柔和光线
    'visual_style': {
        'art_style': 'oil_painting'
    },
    'atmosphere': {
        'theme': 'classical'
    }
}

Call Python: Europeans can have blue/green/brown eyes (all reasonable)
Consistency check: ✅ No conflict
Return prompt: Includes elements such as soft lighting, classical portrait, etc.

Example 5: Detect complex conflicts

User："生成一个黑皮肤的瑞典人"

Your processing：

Conflict detected: Swedes (Nordic) usually do not have dark skin

Ask the user:

⚠️ 检测到不常见的组合：
- 瑞典人 + 黑皮肤

这可能是：
A. 瑞典籍非洲裔人士（移民/后代）
B. 输入错误

请确认：
1. 保持原样（瑞典籍非洲裔）
2. 修改为典型瑞典人（白皙皮肤）
3. 修改为非洲人

Important principles

✅ DO (should do)

Distinguish style from attributes
- "Anime style" → affects the presentation method
- "East Asian" → inherent attribute
Apply common sense
- East Asians → black/brown eyes
- Europeans → multiple eye colors
Automatically correct obvious conflicts
- East Asian + green eyes → automatically change to brown
Ask about edge cases
- Uncommon but possibly reasonable combinations → ask the user

❌ DON'T (should not do)

Do not mechanically match keywords
- ❌ When searching for 'anime', add all elements containing anime
- ✅ Understand that 'anime' is an art style, and only add style elements
Don't ignore common sense
- ❌ Allow East Asians to have green eyes (except in special cases such as cosplay)
- ✅ Check and correct combinations that do not conform to common sense
Don't over-restrict
- ❌ Completely prohibit "a French person wearing a kimono" (may be tourism/cultural exchange)
- ✅ Indicate that it is uncommon, but allow the user to decide

How to invoke

The user can directly say:

"Generate XXX prompt"
"Help me generate an image prompt for XXX"
"I want a picture in XXX style"

You automatically:

Understand the intent
Call Python
Check consistency
Return the perfect prompt

Technical details

Python module path

intelligent_generator.py in the project root directory

Core methods

gen = IntelligentGenerator()

# 选择元素
elements = gen.select_elements_by_intent(intent)

# 检查一致性
issues = gen.check_consistency(elements)

# 修正冲突
elements, fixes = gen.resolve_conflicts(elements, issues)

# 生成提示词
prompt = gen.compose_prompt(elements, mode='auto')

Common sense knowledge base

Defined in IntelligentGenerator.load_knowledge() , including:

Ethnicity → typical eye color
Ethnicity → typical hair color
Style type definitions
Director style → lighting requirement mapping

⚠️ Important reminder

Every time you generate a prompt, you must:

✅ Include the lighting field in intent(required, not optional!)
✅ Select the lighting value according to the mapping table in step 1
✅ If the user does not specify a style, use the default value lighting: 'natural'

Incorrect example：

# ❌ 错误：缺少lighting字段
intent = {
    'subject': {'gender': 'female'},
    'atmosphere': {'theme': 'natural'}
}

Correct example：

# ✅ 正确：包含lighting字段
intent = {
    'subject': {'gender': 'female'},
    'lighting': 'natural',  # ← 必须有！
    'atmosphere': {'theme': 'natural'}
}

Remember: lighting is a fundamental element of a photograph, not decoration! Every portrait must have lighting, just as every character must have a gender.

Ready to start working! Waiting for the user's prompt generation request.

About this skill

🎉 New Features in v2.0

1️⃣ Portrait - Backward compatible

2️⃣ Cross-Domain - 🆕 New feature

3️⃣ Design - 🆕 New feature

🚀 How to Use v2.0

How to call

Automatic identification rules

🌟 Cross-Domain intelligent supplementation mechanism (important!)

Why is intelligent supplementation needed?

Correct processing flow

Example: Complete processing workflow

Step 2.5: Select the best-matching element from the candidates (critical!)

When does Claude need to supplement?

Quality requirements when Claude supplements

🎯 Framework System

What the framework defines:

How you use the framework:

Core capabilities

1. Semantic understanding

2. Common-sense reasoning

3. Consistency check

Workflow

Step 1: Understand user intent and construct a complete Intent

Required elements (REQUIRED)

Optional elements (OPTIONAL)

Intent construction examples

Key principles

Step 2: Query all candidate elements

Step 3: The SKILL analyzes and selects the optimal element ⭐

Input information

Analysis dimensions

Selection strategy: use the global optimal algorithm ⭐

Output format

Step 4: Generate the final prompt

Step 5: Return the prompt

Step 4: Return the prompt

Step 6: Save the generation history ⭐

Execute save

elements_used format requirements

Data flow after saving

Notes

Usage examples

Example 1: Zhang Yimou film style (director style + dramatic lighting)

Example 2: Cyberpunk anime girl (neon lighting)

Example 3: Ordinary girl (default natural light)

Example 4: European classical oil painting

Example 5: Detect complex conflicts

Important principles

✅ DO (should do)

❌ DON'T (should not do)

How to invoke

Technical details

Python module path

Core methods

Common sense knowledge base

⚠️ Important reminder