Core Features
Automatically Activate the Knowledge Base System - This is the core competitive advantage of Novel Writer Skills.
When you mention specific keywords, I will automatically:
- Detect the story's genre, era, and theme
- Load the corresponding writing knowledge base
- Apply professional knowledge throughout the entire creative process
No manual invocation needed - Fully automated, running in the background.
How It Works
Keyword Mapping Table
I monitor the following keywords and automatically activate the corresponding knowledge base:
Genre Knowledge Bases (Genres)
Romance（romance）：
触发词：言情、爱情、恋爱、浪漫、感情线、关系弧、CP、甜文、虐文、
HE、BE、双洁、破镜重圆、先婚后爱、契约关系
激活：templates/knowledge-base/genres/romance.md
Mystery/Detective（mystery）：
触发词：悬疑、推理、侦探、破案、谜团、线索、真相、凶手、犯罪、
密室、诡计、不在场证明、推理小说
激活：templates/knowledge-base/genres/mystery.md
Historical Fiction（historical）：
触发词：历史、古代、朝代、考据、时代背景、历史小说、古言、
穿越、重生古代、架空历史、宫斗、宅斗
激活：templates/knowledge-base/genres/historical.md
Revenge Satisfaction Fiction（revenge）：
触发词：复仇、报仇、打脸、爽文、逆袭、反击、重生复仇、
穿越复仇、系统、金手指、女主爽文、男主爽文
激活：templates/knowledge-base/genres/revenge.md
Martial Arts Fiction（wuxia）：
触发词：武侠、江湖、武功、侠客、门派、武学、剑客、
轻功、内功、武林、江湖恩仇、侠义
激活：templates/knowledge-base/genres/wuxia.md
Reference Libraries (References)
1920s China（china-1920s）：
触发词：1920、民国、军阀、北洋、穿越民国、二十年代、
民国时期、军阀混战
激活：templates/knowledge-base/references/china-1920s/
Automatic Activation Process
Example 1: Single Genre Detection
用户："我要写一部言情小说"
↓
[检测到关键词："言情"]
↓
✓ 自动加载：romance.md
↓
AI回复："太好了！让我帮你创作言情小说。
根据言情类型惯例，我们需要明确几个核心元素...
（自动应用romance.md中的知识）"
Example 2: Multi-Genre Combination Detection
用户："我要写一部1920年代的言情复仇小说"
↓
[检测到关键词："1920"、"言情"、"复仇"]
↓
✓ 自动加载：romance.md
✓ 自动加载：revenge.md
✓ 自动加载：references/china-1920s/
↓
📚 已激活知识库：
- genres/romance.md（言情小说惯例）
- genres/revenge.md（复仇爽文技巧）
- references/china-1920s/（1920年代背景）
↓
AI回复："很好的组合！这是浪漫悬疑+复仇+民国背景。
根据这三个类型的融合，建议...
（同时应用三个知识库的内容）"
Example 3: Continuous Application During the Creative Process
/constitution 阶段：
→ 已激活：romance.md
→ 提醒：言情小说需要HEA/HFN结局承诺
/specify 阶段：
→ 已激活：romance.md + revenge.md
→ 建议：定义关系弧线+复仇目标
/plan 阶段：
→ 已激活：所有知识库
→ 应用：情感节奏点+打脸节奏+1920年代细节
/write 阶段：
→ 已激活：所有知识库
→ 实时：对话技巧+场景描写+时代氛围
/analyze 阶段：
→ 已激活：所有知识库
→ 检查：言情惯例+复仇合理性+历史准确性
Token Efficiency Optimization
Why Is This System Efficient?
Traditional Approach (superpowers-skills mode):
50个独立Skills × 每个40行SKILL.md = 2000 tokens
每次对话都加载所有Skills
Our Approach (Layered Knowledge Base + Detector):
1个detector Skill（本文件）：~100 tokens
按需加载知识库：
- romance.md：~520 tokens
- revenge.md：~480 tokens
- china-1920s/：~650 tokens
总计：100 + 520 + 480 + 650 = ~1750 tokens
（仅加载相关知识库）
Save 75% vs loading 50 irrelevant Skills
Manual Control
View the currently activated knowledge base
Ask at any time:
"当前激活了哪些知识库？"
I will reply:
📚 当前激活的知识库：
✓ genres/romance.md - 言情小说创作惯例
✓ genres/revenge.md - 复仇爽文打脸技巧
✓ references/china-1920s/ - 1920年代中国参考资料
Token消耗：约1650 tokens
Manually Activate a Knowledge Base
If automatic detection fails, you can specify explicitly:
"请加载 romance 和 mystery 知识库"
"这个故事需要 1920s 中国参考资料"
"激活武侠知识库"
Deactivate a Knowledge Base
If a knowledge base is no longer needed:
"停用 revenge 知识库，这个故事不涉及复仇"
Integration with Commands
/constitution - Genre-Specific Principles
Based on the activated knowledge base, provide genre-specific creative principles:
【自动插入到constitution中】
## 类型特定原则（由setting-detector自动添加）
### 言情小说核心原则
1. 关系弧线是主情节
2. HEA/HFN结局是必须的
3. 化学反应真实可信
### 复仇爽文核心原则
1. 仇恨必须深刻具体
2. 打脸节奏渐进式
3. 道德平衡：罪有应得
/specify - Genre-Specific Specifications
Provide the corresponding specification template elements:
【自动提示】
💡 根据已激活的知识库，建议在规格中定义：
【来自romance.md】
- 关系弧线设定
- 情感节奏点规划
- HEA/HFN结局类型
【来自revenge.md】
- 仇恨来源（具体罪行）
- 金手指设定
- 打脸节奏规划
【来自china-1920s】
- 具体年份（1920-1929）
- 地理位置（北京/上海/其他）
- 社会阶层设定
/plan - Genre Knowledge Fusion
Fuse suggestions from multiple knowledge bases into the plan:
【自动应用】
## 章节规划（融合3个知识库）
第1-5章：相遇与仇恨建立
- [romance] 初次相遇场景（meet-cute）
- [revenge] 展示前世/过去的深仇大恨
- [1920s] 民国时代氛围营造
第6-10章：关系升温与小打脸
- [romance] 暧昧期：肢体接触、情感积累
- [revenge] 第一次打脸：仇人惊讶
- [1920s] 服饰、礼仪、社会规则展现
...
/write - Real-Time Writing Guidance
Automatically apply relevant knowledge while writing:
【写作前自动提醒】
📋 本章应用的知识库：
✓ romance.md - 对话火花技巧
→ 调情对话：挑战而非顺从
→ 肢体语言：眼神、触碰、距离
✓ revenge.md - 打脸场景爽感
→ 期待 → 反转 → 惊愕 → 霸气
✓ 1920s - 时代细节
→ 称呼：大人、老爷、小姐
→ 场景：茶馆、洋行、租界
/analyze - Genre Convention Checks
Perform the corresponding checks based on the activated knowledge base:
【自动分析】
## 类型符合度分析
### 言情小说检查（基于romance.md）
- [x] 关系弧线是主情节
- [x] 包含必备情感节奏点
- [ ] ⚠️ 缺少初吻场景（建议在第15章）
- [x] HEA结局承诺明确
### 复仇爽文检查（基于revenge.md）
- [x] 仇恨深刻具体
- [x] 打脸频率合理（每5章1次）
- [x] 主角实力提升合理
### 历史准确性检查（基于1920s）
- [x] 时代背景正确
- [ ] ⚠️ 第8章出现"手机"（时代不符）
- [x] 称呼系统正确
Knowledge Base Extension
Currently Supported Knowledge Bases
| Category | Completed | Planned |
|---|---|---|
| Genre Knowledge | 5 | 10+ |
| Reference Materials | 1 | 20+ |
Completed（v1.0）：
- genres/romance.md
- genres/mystery.md
- genres/historical.md
- genres/revenge.md
- genres/wuxia.md
- references/china-1920s/
Planned（v1.1+）：
- genres/fantasy.md (Fantasy)
- genres/sci-fi.md (Science Fiction)
- genres/horror.md (Horror)
- references/ancient-china/ (Various Dynasties)
- references/modern-workplace/ (Modern Workplace)
Adding a New Knowledge Base
- In
templates/knowledge-base/Create a file in the corresponding directory - Update
templates/knowledge-base/README.md's keyword mapping table - setting-detector will automatically recognize it (no need to modify this file)
Example: Adding a Fantasy Knowledge Base
# 1. 创建文件
touch templates/knowledge-base/genres/fantasy.md
# 2. 更新README.md关键词映射
fantasy:
keywords: [奇幻, 魔法, 世界构建, 魔法系统]
auto_load: genres/fantasy.md
# 3. 完成！下次用户说"奇幻"就会自动激活
Intelligent Features
1. Fuzzy Matching
It can recognize even if it's not an exact keyword:
"我想写个穿越到民国的复仇故事"
↓
识别："民国" → 1920s中国
"复仇" → revenge
"女主重生后要报仇"
↓
识别："重生" + "报仇" → revenge（重生复仇）
2. Context Understanding
Continuously recognize based on the conversation context:
第1条消息："我要写小说"
→ 未激活任何知识库（等待更多信息）
第2条消息："主角是侦探"
→ 激活mystery.md
第3条消息："还有感情线"
→ 额外激活romance.md
→ 当前激活：mystery + romance（浪漫悬疑）
3. Automatic deduplication
Avoid duplicate activation:
用户："这是武侠小说，江湖背景，有武功"
↓
识别到3个武侠关键词，但只激活1次wuxia.md
Frequently Asked Questions
Q: How do I know which knowledge base has been activated?
A: Just ask me at any time "Which knowledge bases are currently activated?" and I will list them.
Q: What should I do if automatic detection is wrong?
A: Just tell me directly:
"这不是言情小说，请停用romance知识库"
"这是科幻小说，请激活sci-fi知识库"（如果已有）
Q: How many knowledge bases can be activated at the same time?
A: Theoretically unlimited, but 3-5 is recommended as optimal:
- Too few: guidance is not comprehensive enough
- Too many: high token consumption, possible conflicts
Practical experience: 3 knowledge bases (such as romance + revenge + 1920s) are already very rich.
Q: Will knowledge bases conflict?
A: No! Knowledge bases are designed to work together:
- romance + mystery = romantic suspense✓
- fantasy + romance = fantasy romance✓
- historical + revenge = ancient revenge✓
Q: When should I manually specify a knowledge base?
A: When:
- Highly specialized terminology is used (I may not have recognized it)
- You want to try a different type of fusion
- Automatic detection does not match your intent
Q: Will knowledge bases stay activated all the time?
A: Yes, until:
- You explicitly say "deactivate XX knowledge base"
- Or a new conversation starts
- Or you switch to a completely different story
Best practices
1. Clarify the genre early in the creative process
推荐做法：
用户："我要写一部民国背景的复仇言情小说"
不推荐：
用户："我要写小说"
AI："好的"
用户："主角姓李"
AI："嗯"
（20轮对话后才提到是什么类型）
2. Regularly check activation status
Ask once every 5-10 chapters:
"当前激活的知识库还适合吗？"
I will check and suggest whether adjustments are needed.
3. Genre fusion should be reasonable
✓ 好的组合：
- romance + mystery（浪漫悬疑）
- historical + romance（古言）
- revenge + romance（复仇+感情）
⚠️ 困难的组合：
- horror + romance（恐怖言情？读者可能不适应）
- mystery + wuxia（推理武侠，需要特殊设计）
不是不能做，但需要特别小心融合方式。
4. Make use of reference materials
If the story has a clear historical background, be sure to activate the corresponding reference materials:
"1920年代" → china-1920s
"古代" → 指定具体朝代（如"唐代"、"明朝"）
Historical details are accurate, and readers' immersion is strong.
Technical notes
Detection algorithm
1. 用户输入 → 提取关键词
2. 关键词 → 映射到知识库（基于README.md）
3. 去重 → 避免重复激活
4. 加载 → Read相应的.md文件
5. 应用 → 在后续对话中持续应用知识
Token management
激活状态检查：每次对话开始时
知识库内容：按需加载，缓存在对话上下文中
更新策略：只在明确需要时重新读取
Collaboration with other Skills
setting-detector is the "knowledge dispatcher", and other Skills are the "executors":
setting-detector → 决定加载哪些知识
consistency-checker → 基于激活的知识库检查一致性
workflow-guide → 根据类型提供特定流程指导
genre-specific skills → 提供类型特定的深度技巧
Summary
setting-detector is the Novel Writer Skillsbrain：
✓ Automatically detects story settings ✓ Intelligently activates the corresponding knowledge bases ✓ Token-efficient (loads on demand) ✓ Continuously applies professional knowledge ✓ Requires no user intervention
Let AI automatically become an expert in your story's genre! 🧠✨
This Skill version: v1.0 Last updated: 2025-10-18 Dependencies: templates/knowledge-base/ system Collaboration: consistency-checker, workflow-guide, genre-specific skills