playbooks

home / skills / wordflowlab / novel-writer-skills / requirement-detector

requirement-detector skill

/templates/skills/quality-assurance/requirement-detector

This skill detects your writing requirements and loads the matching guidelines to tailor your novel with the chosen style rules.

npx playbooks add skill wordflowlab/novel-writer-skills --skill requirement-detector

Review the files below or copy the command above to add this skill to your agents.

Files (4)
SKILL.md
4.7 KB
---
name: requirement-detector
description: "探测用户的写作规范需求并加载对应文档。当用户提到AI味重、去AI味、自然、爽文、快节奏、爽点、无毒点、不降智、严肃文学、有深度、强情绪、打动人、甜文、撒糖、虐文、虐心、BE等关键词时自动激活。适用于讨论写作要求、AI去味方法、节奏控制、情感表达时使用。"
allowed-tools: Read, Edit
---

# 写作规范探测器

## 核心功能

自动探测用户的写作规范需求，并无缝加载对应的规范知识库。

## 可探测的规范

### 1. anti-ai-v4（强力AI去味）
- **适合**：对AI味零容忍的作品、需要极致口语化的小说
- **特点**：200+ 禁用词汇表、6层规则体系、极致自然化
- **触发词**：AI味重、去AI味、太机器、不自然、最强去味

### 2. anti-ai-v3（标准AI去味）
- **适合**：大部分现代小说、需要平衡自然感和文学性
- **特点**：平衡的去AI策略、适度的词汇限制
- **触发词**：自然一点、少一点AI感、适度去AI、标准去味

### 3. fast-paced（快节奏/爽文规范）
- **适合**：都市爽文、玄幻升级流、系统文、重生文
- **特点**：每章至少2个爽点、打脸/装逼/升级公式
- **触发词**：爽文、快节奏、爽点、打脸、装逼、升级

### 4. no-poison（无毒点规范）
- **适合**：所有类型小说（通用）、重视逻辑性的作品
- **特点**：避免降智、强行误会、玛丽苏、工具人、逻辑漏洞
- **触发词**：无毒点、不降智、逻辑自洽、合理、不强行

### 5. serious-literature（严肃文学规范）
- **适合**：现实主义小说、历史小说、社会问题小说
- **特点**：现实主义手法、思想性和哲学探讨、语言克制
- **触发词**：严肃文学、纯文学、有深度、思想性、人性探讨

### 6. strong-emotion（强情绪规范）
- **适合**：情感类小说、虐文、需要感染力的作品
- **特点**：每10-15章至少1个情绪爆发点、波浪式情绪节奏
- **触发词**：情绪饱满、情感强烈、有感染力、打动人心

### 7. romance-sweet（甜文规范）
- **适合**：言情甜文、现代甜宠、轻松向恋爱故事
- **特点**：糖分密度配置、冲突控制、禁止BE倾向
- **触发词**：甜文、撒糖、高糖、齁甜、甜宠、HE

### 8. romance-angst（虐文规范）
- **适合**：虐恋小说、悲剧向爱情、刀子文
- **特点**：虐要有理有度、虐点设计、虐度等级控制
- **触发词**：虐文、虐心、BE、刀子、分离虐、时机虐

详细关键词列表和触发场景请参阅 [KEYWORDS.md](KEYWORDS.md)。

---

## 激活后的工作流程

### 步骤 1：确认规范选择

当探测到关键词后，首先确认用户意图：

```
我注意到你提到了【AI味重、去AI味】，这通常对应 anti-ai-v4 规范。

我可以：
1. 📖 展示这个规范的详细文档
2. ✅ 直接应用这个规范到你的项目
3. 🔍 看看其他规范选项

你想要哪一个？
```

### 步骤 2：加载规范知识库

如果用户确认，使用 Read 工具读取对应的知识库文件：

```
📖 读取：.claude/knowledge-base/requirements/{requirement-name}.md
```

然后展示核心要点，完整规范已加载到上下文中。

### 步骤 3：写入配置标记

使用 Edit 工具在用户的 `specification.md` 中添加配置：

```yaml
---
writing-requirements:
  - anti-ai-v4
  - fast-paced
---
```

提示用户配置已保存，在执行 /write 时会自动遵守这些规范。

---

## 多规范处理

当探测到多个规范关键词时，进行兼容性检测：

- **兼容组合**：自动组合并说明综合效果
- **冲突组合**：提示矛盾并给出解决方案

详细的冲突处理逻辑请参阅 [CONFLICT_RESOLUTION.md](CONFLICT_RESOLUTION.md)。

---

## 主动建议

- **基于风格建议规范**：如果 style-detector 已激活，主动推荐配套规范
- **中途调整支持**：持续监控用户反馈，支持规范调整

---

## 与 style-detector 的配合

本 Skill 专注于规范层面，与 style-detector（文风层）协同工作：

```
用户输入："我要写都市爽文，口语化，去AI味"

style-detector → 识别：natural-voice
requirement-detector → 识别：fast-paced + anti-ai-v4

最终配置：
writing-style: natural-voice
writing-requirements:
  - fast-paced
  - anti-ai-v4

效果：自然人声的快节奏爽文，强力去AI味
```

---

## 使用示例

完整的对话示例和成功案例请参阅 [EXAMPLES.md](EXAMPLES.md)。

---

**Skill 版本**：v1.1
**最后更新**：2025-10-19
**兼容版本**：novel-writer-skills v1.0.5+

Overview

This skill detects a user's writing requirement keywords and automatically loads the matching requirement document into the session. It activates when users mention terms like AI flavor, de-AI, natural, fast-paced, sweet, angst, no-toxic-points, serious literature, or strong emotion, and helps enforce the selected constraints during writing.

How this skill works

The skill scans user input for a curated keyword list and maps matches to predefined requirement profiles (e.g., anti-ai-v4, fast-paced, no-poison). Once a profile is suggested, it can display the requirement summary, load the full knowledge base into context, and write configuration markers into the project spec so future writing tasks follow those rules. When multiple profiles are detected, the skill checks compatibility and proposes merges or conflict resolutions.

When to use it

  • When you want to remove or reduce AI-generated phrasing from prose (de-AI).
  • When planning a fast-paced or '爽文' narrative that needs chapter-level hooks.
  • When you need to ensure logical consistency and avoid 'toxic' plot devices.
  • When writing emotionally intense or cathartic scenes and need rhythm control.
  • When producing romance that must be strictly HE (sweet) or BE/angst (hurt).

Best practices

  • Confirm the detected requirement before applying it to avoid unwanted constraints.
  • Combine compatible profiles (e.g., natural-voice + fast-paced) and review suggested trade-offs.
  • Use the loaded requirement summaries to set clear chapter-level goals (pacing, emotion beats).
  • Adjust requirements mid-project based on reader feedback; the skill supports iterative updates.
  • When conflicts arise, prefer explicit resolution suggestions from the skill rather than silent merging.

Example use cases

  • User says 'make it less AI-sounding' → suggest anti-ai-v4 and offer to load that guide.
  • User requests 'fast-paced, lots of hits per chapter' → attach fast-paced rules and pacing checklist.
  • Writer building a romance: 'I want high sugar, HE' → apply romance-sweet and block BE elements.
  • Draft has logic holes → apply no-poison profile to flag and remove downgraded-intelligence beats.
  • Planning a tragic arc: 'I need controlled angst, not melodrama' → apply romance-angst with degree limits.

FAQ

What if multiple conflicting requirements are detected?

The skill runs a compatibility check, highlights conflicts, and offers safe resolution options or priority choices you can accept.

Can I change requirements mid-project?

Yes. The skill supports updating the spec and will reapply the active requirement set to future writes.

Does it modify my text automatically?

It loads rules and writes configuration markers; you choose whether to apply the rules automatically during /write operations.