release: publish 0.2.0

2026-03-21 01:55:08 +08:00
1 parent f98538ff1d
commit cd146dea89
@@ -1,25 +1,47 @@
-# Changelog
+# 变更日志

-All notable changes to `novel-control-station` will be documented in this file.
+`novel-control-station` 的重要变更都会记录在这个文件中。
+
+## 0.2.0 - 2026-03-21
+
+### 新增
+
+- 项目级章节标题控制，包括命名体系锁定、章节控制卡中的标题候选，以及成稿后的题章贴合度复核。
+- marathon 启动交接流程，包括 `assets/codex-continue-novel.ps1` 模板和项目根目录 runner 生成。
+- 章节标题、质量/回写复核、读者留存失败模式、marathon 启动交接等补充控制参考文档。
+- 扩展后的内部风格模块参考集，覆盖幽默、悬念、推理、爱情、恐怖、奇幻和文学风格。
+
+### 变更
+
+- `README.md` 现已记录章节标题流程、内部风格模块调度，以及 marathon runner 启动说明。
+- `SKILL.md` 现已将章节标题、章节回写和 marathon handoff 纳入一等工作流阶段。
+- `references/document-templates.md` 现已记录章节标题设置，以及激活主风格的兑现预期。
+- 内部风格模块已从简短路由说明改写为可操作的创作参考文档，补齐了子类型覆盖、修订检查和更柔性的自适应约束。
+
+### 移除
+
+- 发布树中的 `tests/` 目录及其测试脚本。
+- 发布树中的 `evals/` 评测目录及场景文件。
+- 本次开发阶段生成的 `docs/plans/` 实施计划草稿。

 ## 0.1.0 - 2026-03-19

-Initial public release.
+首次公开发布。

-### Added
+### 新增

- Single-entry control workflow for Chinese long-form fiction planning, drafting, repair, and continuation.
- Document-driven novel architecture with standard project files, dynamic state updates, and audit logging.
- Full-outline-first workflow with interview, alignment, cast dossier, and chapter-by-chapter progression.
- Multi-line structure control for main plot, subplots, relationships, foreshadowing, and world-rule continuity.
- Built-in secondary recall layer for graph-style retrieval without replacing the canonical project files.
- Chapter control card, forgotten-element monitoring, line-heat checks, and anti-drift continuity safeguards.
- Style-dispatch design for suspense, humor, romance, literary, historical, and other genre-specific references.
- Authenticity and de-AI pass focused on preserving Chinese prose texture, character voice, and narrative identity.
- Marathon drafting mode that continues chapter by chapter until the planned ending is naturally reached.
+- 面向中文长篇小说规划、创作、修补和续写的单入口控制工作流。
+- 以标准项目文件、动态状态更新和审计日志为核心的文档驱动小说架构。
+- 先完整大纲、后正文的工作流，包括访谈、对齐、人物档案和逐章推进。
+- 面向主线、支线、关系线、伏笔线和世界规则连续性的多线结构控制。
+- 内建图式召回层，用于复杂项目的图结构检索，但不替代规范项目文件。
+- 章节控制卡、遗忘元素监测、线热度检查和防漂移连续性保护。
+- 面向悬念、幽默、爱情、文学、历史等题材参考的风格调度设计。
+- 以保留中文质感、人物声音和叙事身份为目标的真实性与去 AI 修订流程。
+- 可逐章持续推进，直到按既定大纲自然收束的 marathon 创作模式。

-### Packaging
+### 打包

- Public-facing `README.md` with product-style positioning.
- `LICENSE` under MIT.
- `RELEASE_NOTES.md` for release overview and usage guidance.
+- 面向公开使用者的 `README.md`，采用产品式定位说明。
+- 基于 MIT 的 `LICENSE`。
+- 用于发布概览和使用提示的 `RELEASE_NOTES.md`。
@@ -3,10 +3,10 @@
 <div align="center">
  <p><strong>把中文长篇小说创作，从一次性生成，升级成可控、可追踪、可持续推进的创作系统。</strong></p>
  <p>
-    <img alt="License: MIT" src="https://img.shields.io/badge/license-MIT-111111">
-    <img alt="Chinese Long-form Fiction" src="https://img.shields.io/badge/focus-Chinese%20Long--Form%20Fiction-0f766e">
-    <img alt="Doc-Driven" src="https://img.shields.io/badge/architecture-Doc--Driven-1d4ed8">
-    <img alt="Continuity Control" src="https://img.shields.io/badge/control-Continuity%20Engine-7c3aed">
+    <img alt="许可证: MIT" src="https://img.shields.io/badge/license-MIT-111111">
+    <img alt="聚焦: 中文长篇小说" src="https://img.shields.io/badge/focus-Chinese%20Long--Form%20Fiction-0f766e">
+    <img alt="架构: 文档驱动" src="https://img.shields.io/badge/architecture-Doc--Driven-1d4ed8">
+    <img alt="控制: 连续性引擎" src="https://img.shields.io/badge/control-Continuity%20Engine-7c3aed">
  </p>
 </div>

@@ -14,8 +14,24 @@

 它会先对齐题材、受众、主角、规模与结局方向，再生成完整大纲和人物档案；进入正文阶段后，又会按章节读取项目文档、控制结构推进、回写动态状态、修正连续性风险，并在需要时调用内置风格与技法文档，持续压制设定漂移、人物失真、关系断档、伏笔遗忘和文风塌陷。

+当前正式版本 `0.2.0` 进一步补强了三个升级面：
+
+- 章节标题控制已经并入章节流程，可在项目级锁定命名体系、在单章生成候选标题并在成稿后复核题章贴合度
+- 内置风格模块已经整理为选择性加载的内部文档系统，覆盖幽默、悬念、推理、爱情、恐怖、奇幻、文学等风格
+- 连续创作模式新增项目根目录启动脚本模板，可为 marathon / 疯狂创作交接准备 `codex-continue-novel.ps1`
+
 如果你想要的是一个能从开书、控盘、扩写、修补一路推进到自然完结的中文小说生产中枢，这个技能就是为这个目标设计的。

+## Star 历史
+
+<a href="https://www.star-history.com/?repos=jingtai123%2FNovel-Control-Station-Skill&type=date&legend=top-left">
+ <picture>
+   <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/image?repos=jingtai123/Novel-Control-Station-Skill&type=date&theme=dark&legend=top-left" />
+   <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/image?repos=jingtai123/Novel-Control-Station-Skill&type=date&legend=top-left" />
+   <img alt="Star 历史图表" src="https://api.star-history.com/image?repos=jingtai123/Novel-Control-Station-Skill&type=date&legend=top-left" />
+ </picture>
+</a>
+
 ## 它解决什么问题

 长篇小说最难的，通常不是“第一章写不出来”，而是：
@@ -99,7 +115,7 @@
 它吸收了图结构管理角色、事件、关系、伏笔、世界规则的思路，但不会制造第二套互相打架的真相源。  
 标准项目文件始终是唯一真值层，图式回忆只作为临时控制视图，用于快速提取本章必须回读和处理的内容。

-### 7. 章节控制与强钩子闭环
+### 7. 章节控制、章节命名与强钩子闭环

 每一章在动笔前，都会先明确：

@@ -109,19 +125,24 @@
 - 本章冲突是什么
 - 结尾留下什么余波或牵引

+如果项目启用了章节名，它还会把章节命名纳入控制流程，先锁定全书命名体系，再为单章生成多个候选名，最后在成稿后复核标题是否仍然贴合本章真正的重心、悬念和余波。
+
 如果一章只是把人物从 A 地送到 B 地，它会把这类章节视为结构性疲软。

 ### 8. 风格调度与按需引用

 技能已经为不同风格和题材预留了独立参考文档，不会把所有技法硬塞进同一层提示。  
-在需要时，它可以按场景调用不同的写法逻辑，例如：
+在需要时，它会通过 `references/style-modules/index.md` 选择性加载内部风格模块，先读 `core.md`，只有章节真的需要时才继续展开更深文档。当前内置模块包括：

- 悬疑
 - 幽默
- 言情
- 历史
- 文学化表达
- 群像叙事
+- 悬念
+- 推理
+- 爱情
+- 恐怖
+- 奇幻
+- 文学
+
+它支持主风格与辅助风格协同，但不会把所有风格一次性压进同一章。

 这意味着它不是单一文风生成器，而是带有调度能力的风格中枢。

@@ -135,13 +156,14 @@

 目标不是更平，而是更像这本小说自己该有的语言。

-### 10. 疯狂创作模式与自然完结
+### 10. 疯狂创作模式、启动脚本与自然完结

 用户确认大纲和人物档案后，可以开启连续创作模式。  
 在这个模式里，技能不会在每一章后反复请求人工确认，而是持续执行：

 - 读取项目文档
 - 生成章节控制卡
+- 生成并复核章节标题（如果项目使用章节名）
 - 起草正文
 - 进行逻辑校验与必要重写
 - 更新动态状态
@@ -150,17 +172,24 @@

 直到按完整大纲自然完结为止。

+如果用户要开启疯狂创作 / 自动续写模式，技能还会先准备项目根目录启动脚本：
+
+- 从 `assets/codex-continue-novel.ps1` 生成项目专用 `codex-continue-novel.ps1`
+- 替换项目根目录占位符
+- 在交接时明确给出启动命令：
+  `powershell -ExecutionPolicy Bypass -File .\codex-continue-novel.ps1`
+
 ## 能力看板

 | 模块 | 能力表现 |
 |---|---|
 | 立项层 | 题材定位、受众识别、主角与冲突对齐、规模判断 |
 | 设计层 | 完整大纲、人物档案、世界规则、关系结构、多线编排 |
-| 创作层 | 章节控制、钩子设计、闭环推进、场景落地、节奏调度 |
+| 创作层 | 章节控制、章节命名、钩子设计、闭环推进、场景落地、节奏调度 |
 | 连续性层 | 设定防漂移、人物防失真、关系防断档、伏笔防遗失 |
-| 调度层 | 风格参考按需读取、题材规则按需触发、复杂项目回忆切片 |
+| 调度层 | 风格模块按需读取、题材规则按需触发、复杂项目回忆切片 |
 | 修订层 | 去 AI 味、真实性增强、人物声音校准、叙事力度修正 |
-| 项目层 | 动态状态管理、日志留痕、长期推进、自然完结控制 |
+| 项目层 | 动态状态管理、标题体系锁定、日志留痕、长期推进、自然完结控制 |

 ## 它是怎么工作的

@@ -177,16 +206,19 @@
 ### 第三步：按文档驱动章节创作

 每一章开始前，先读取项目真值文件；必要时再调用回忆切片、风格文档、控制卡文档。  
+如果项目使用章节标题，还会在章节控制卡阶段先生成候选标题并锁定工作标题。
 章节不是“随机生成”，而是“在可控上下文内推进”。

 ### 第四步：写完一章，项目状态立刻更新

-章节完成后，技能会更新动态状态与关键记录，避免前账失踪、人物弧光断裂、情绪债务蒸发、伏笔遗忘。
+章节完成后，技能会更新动态状态与关键记录，必要时回写章节标题、路线图和风格指南，避免前账失踪、人物弧光断裂、情绪债务蒸发、伏笔遗忘。

 ### 第五步：跑完整本，而不是只写局部精彩

 它的终极目标不是把单章写漂亮，而是让整部中文长篇小说从立项到完结都维持稳定质量与结构压力。

+如果要切入 marathon handoff，它还会先确认启动脚本已经写入项目根目录，避免“想自动继续，但下次不知道怎么恢复”。
+
 ## 为什么它和常见小说技能不一样

 ### 它不是一次性长提示词
@@ -199,6 +231,11 @@
 风格、结构、连续性、修订、章节控制都被彻底合并为一组内置文档，由总控入口统一调度。  
 用户不需要在多个技能之间切来切去争夺上下文。

+### 它不是把章节标题和风格当装饰层
+
+章节标题、风格模块、章节验收、动态回写都已并入主流程。
+它不会把标题当成事后贴标签，也不会把风格当成统一撒味精。
+
 ### 它不是只会持续推进

 除了持续写下去，它还擅长：
@@ -241,15 +278,21 @@ novel-control-station/
 ├── LICENSE
 ├── CHANGELOG.md
 ├── RELEASE_NOTES.md
-├── evals/
+├── assets/
+│   └── codex-continue-novel.ps1
 └── references/
+    ├── bootstrap-and-marathon-handoff.md
    ├── document-templates.md
    ├── interview-and-handoff-flow.md
    ├── chapter-control-card.md
+    ├── chapter-title-method.md
    ├── continuity-and-marathon-mode.md
+    ├── quality-and-writeback-checks.md
+    ├── reader-retention-and-ai-failure-modes.md
    ├── graph-and-recall-control.md
    ├── forgotten-elements-and-line-heat.md
    ├── authenticity-and-de-ai-pass.md
+    ├── style-modules/
    ├── ...
 ```

@@ -259,8 +302,10 @@ novel-control-station/
  - 总控入口，定义整体流程、规则和触发机制。
 - `references/`
  - 内置文档库，按需读取，不一次性全量展开。
- `evals/`
-  - 用于后续做技能压力测试、规则验证和行为评估。
+- `references/style-modules/`
+  - 内部风格模块索引与细分文档，按章节选择性加载。
+- `assets/`
+  - 项目启动脚本模板与其他可复用资源。

 ## 快速开始

@@ -279,21 +324,28 @@ novel-control-station/
 你先补齐必要问题，整理成标准文档，再进入连续创作模式，按完整大纲自然完结停。
 ```

+```text
+使用 novel-control-station。
+这本书的章节需要标题。请先帮我确定全书章节命名体系，再在每章控制卡里给出 3 到 5 个候选标题，并在正文成稿后复核定稿标题。
+```
+
 ## 推荐使用方式

-1. 先交代题材、受众、主角、篇幅、结局倾向。
+1. 先交代题材、受众、主角、篇幅、结局倾向，以及你是否希望章节使用标题。
 2. 先让技能生成完整大纲和人物档案。
 3. 对大纲和人物做一次集中修改。
-4. 确认后再开启连续创作模式。
+4. 如果项目使用章节标题，先锁定标题体系。
+5. 如果要进入连续创作模式，先让技能为项目生成根目录启动脚本。
+6. 确认后再开启连续创作模式。

 这样它的控盘能力、连续性表现和中后期稳定性会明显更强。

 ## 开源信息

- License: MIT
- Default output language: Chinese
- Positioning: 面向中文长篇小说项目的控制中枢型技能
- Release type: 首个公开发布版本
+- 许可证：MIT
+- 默认输出语言：中文
+- 定位：面向中文长篇小说项目的控制中枢型技能
+- 当前正式版本：`0.2.0`

 ## 一句话总结

@@ -1,61 +1,100 @@
-# Release Notes
+# 发布说明

-## Novel Control Station 0.1.0
+## Novel Control Station 0.2.0 版本

-`novel-control-station` is now packaged as a public open-source skill for Chinese long-form fiction work.
+`0.2.0` 是在 `0.1.0` 基础上的一次功能型升级发布。它把此前已经接入源码树、但还未正式打包说明的能力，整理成了清晰可用的正式版本。

-This first release positions the skill as a novel control system rather than a one-shot writing prompt. Its goal is to help users align a book before drafting, keep structure and character logic stable during drafting, and sustain long-form output until the story reaches its planned ending.
+### 本版本新增并正式发布的内容

-## Why this release matters
+- 章节标题工作流：支持项目级命名体系、逐章候选标题生成、工作标题选择以及成稿后的题章贴合度复核。
+- 基于 `assets/codex-continue-novel.ps1` 生成项目根目录 runner 的 marathon 启动交接流程。
+- 位于 `references/style-modules/` 下的内部风格模块系统，覆盖幽默、悬念、推理、爱情、恐怖、奇幻和文学风格。
+- 已整合的章节验收与回写参考，包括自适应降权、原创性约束和读者留存失败模式。

-Most AI novel workflows can produce scenes, but they struggle to run a full-length Chinese novel without drift. The breakdown usually appears in the middle and late stages: settings blur, relationships flatten, foreshadowing disappears, and chapter rhythm starts serving output volume instead of story pressure.
+### 为什么这次升级重要

-This release is built to address that exact failure pattern with:
+这些升级进一步把技能从“只会靠提示词写正文”推进到“能管理小说项目”的状态：

- interview-first alignment
- full-outline-first design
- document-driven project truth
- chapter control before drafting
- dynamic state updates after each chapter
- continuity and recall safeguards for long projects
+- 章节标题不再是装饰性标签，而是受控的叙事信号
+- 风格指导不再全量混读，而是按章节选择性加载
+- marathon 续写不再依赖用户记住恢复命令，而是有显式 handoff 机制
+- 章节验收与回写规则对自适应降权和失败检测都写得更明确

-## Who this is for
+### 发布树整理

- Users writing Chinese web novels or serialized fiction
- Users building multi-plot, cast-dense, foreshadow-heavy long-form fiction
- Users who need AI to keep writing beyond the first few strong chapters
- Users who want stronger project control, not just prettier sample paragraphs
+- 已移除 `tests/` 目录及测试脚本
+- 已移除 `evals/` 评测目录与场景文件
+- 已移除本次开发阶段产生的 `docs/plans/` 实施草稿

-## Standout capabilities
+这样当前发布树会更聚焦于技能本体、参考文档和运行资源，不再混入测试/计划产物。

- Run a novel like a managed project, not a temporary chat thread.
- Keep characters, relationships, plotlines, promises, and emotional debt in active circulation.
- Switch from planning to marathon drafting without dropping internal control.
- Use style references on demand instead of flattening every genre into one generic tone.
- Apply de-AI cleanup without erasing voice, rhythm, or Chinese prose texture.
- Support repair work when a novel has already started drifting or losing pressure.
+### 本版本应重点阅读的文档

-## Included in this release
+使用 `0.2.0` 时，建议优先阅读：

- Core skill entry in `SKILL.md`
- Public project overview in `README.md`
- Reference documents under `references/`
- Evaluation scaffolding under `evals/`
- MIT license and release metadata
+- `README.md`：产品级定位与工作流概览
+- `SKILL.md`：当前生效的总控流程
+- `references/document-templates.md`：项目真值文件结构
+- `references/style-modules/`：选择性风格加载
+- `references/bootstrap-and-marathon-handoff.md`：自动续写启动交接
+- `references/quality-and-writeback-checks.md`：章节验收与回写细则

-## Known limits
+## Novel Control Station 0.1.0 版本

- This skill is optimized for Chinese fiction workflows, not general-purpose creative writing.
- It is intentionally heavier than lightweight prompt-based story generation.
- Final quality still depends on the user's premise, revision standards, and willingness to align the outline before drafting.
+`novel-control-station` 已作为一个面向中文长篇小说工作的公开开源技能完成打包发布。

-## Recommended usage
+首个版本将这个技能定位为“小说控制系统”，而不是“一次性写作提示词”。它的目标是在正式写正文之前先帮助用户把书的方向对齐，在创作过程中维持结构和人物逻辑稳定，并持续推动长篇项目直到抵达计划中的结尾。

-1. Start with the novel's genre, audience, protagonist, core personality, intended length, and ending direction.
-2. Let the skill ask follow-up questions until the plot direction is aligned.
-3. Review the complete outline and cast dossier before authorizing drafting.
-4. Use continuous drafting only after the control documents are stable.
+## 为什么这个版本重要

-## License
+多数 AI 小说工作流都能写出局部场景，但很难在一部长篇中文小说里长期维持不漂移。问题通常出现在中后段：设定开始模糊，关系开始变平，伏笔开始失踪，章节节奏也开始为“产字数”服务，而不是为故事压力服务。

-Released under the MIT License.
+这个版本正是围绕这些失控模式来设计的，核心手段包括：
+
+- 先访谈、后对齐
+- 先完整大纲、后正文
+- 文档驱动的项目真值层
+- 动笔前先做章节控制
+- 每章之后立即更新动态状态
+- 面向长篇项目的连续性与召回防护
+
+## 适合谁用
+
+- 写中文网文或连载小说的用户
+- 正在构建多主线、人物密集、伏笔密集的长篇项目的用户
+- 希望 AI 不只会写前几章，而能持续往后推进的用户
+- 想要更强项目控制，而不只是更漂亮样章的用户
+
+## 突出能力
+
+- 把小说当成可管理项目来运行，而不是临时会话线程。
+- 让人物、关系、情节线、承诺和情感债持续处于激活状态。
+- 从规划阶段切换到 marathon 创作时，不丢失内部控制。
+- 按需调用风格参考，而不是把所有类型压扁成同一种通用语气。
+- 做去 AI 修订时不抹掉人物声音、句法节奏和中文质感。
+- 当小说已经开始跑偏或失压时，支持中途修补。
+
+## 0.1.0 首发内容
+
+- 位于 `SKILL.md` 的核心技能入口
+- 位于 `README.md` 的公开项目说明
+- `references/` 下的参考文档
+- 首发阶段附带的 `evals/` 评测脚手架（已在 `0.2.0` 发布树中移除）
+- MIT 许可证和发布元数据
+
+## 已知限制
+
+- 这个技能针对的是中文小说工作流，不是通用创意写作。
+- 它刻意比轻量提示词式故事生成更重。
+- 最终质量仍然取决于用户的题材设定、修订标准，以及是否愿意先把大纲对齐。
+
+## 推荐使用方式
+
+1. 先给出小说题材、受众、主角、核心性格、目标篇幅和结局方向。
+2. 让技能继续追问，直到整体走向对齐。
+3. 在授权正文创作前，先审核完整大纲和人物档案。
+4. 只有在控制文档稳定后，再进入连续创作。
+
+## 许可证
+
+基于 MIT 许可证发布。
@@ -1,6 +1,6 @@
 ---
 name: novel-control-station
-description: Use when writing, planning, continuing, repairing, or revising Chinese long-form fiction with recurring characters, multiple plotlines, persistent world rules, chapter-by-chapter continuity needs, or style-specific constraints.
+description: Use when writing, planning, continuing, repairing, revising, or running marathon/"疯狂写作"/auto continuation mode for Chinese long-form fiction with recurring characters, multiple plotlines, persistent world rules, chapter-by-chapter continuity needs, or style-specific constraints.
 ---

 # Novel Control Station
@@ -36,6 +36,8 @@ Do not use this skill for one-off poems, short jokes, or isolated scenes that do
 - Do not satisfy forgotten-element checks with token cameos, cosmetic mentions, or checklist references. Re-entry must change pressure, debt, or expectation.
 - Do not run de-AI cleanup as blind flattening. Preserve genre register, era texture, narrator stance, and character voice.
 - Do not force scene ladders into rigid formula when the chapter needs looser movement. Use scene control to preserve pressure, not to fake architecture.
+- Do not use chapter titles as spoiler summaries, empty riddles, or decorative labels detached from chapter pressure.
+- If the project uses chapter titles, lock a naming system at project level and keep title voice consistent unless the book is intentionally entering a new phase.
 - If information is missing or contradictory, explain the risk and let the user choose whether to refine details or draft directly.
 - If drafting proceeds with assumptions, record them in the dynamic state file.
 - After every chapter, update dynamic state before moving on.
@@ -59,6 +61,28 @@ Maintain these files for every novel project:

 Read [document-templates.md](references/document-templates.md) when creating or restoring these files.

+## Startup Project Bootstrap
+
+When starting a new novel project, treat the project root as an operational control surface, not just a folder that stores markdown files.
+
+After creating the standard files, also create:
+
+- `codex-continue-novel.ps1` in the project root
+
+This script is the looping continuation entry point for marathon mode. It should be ready before the project leaves startup.
+Use the exact template in [assets/codex-continue-novel.ps1](assets/codex-continue-novel.ps1) and replace only the project root placeholder before writing the file into the project root.
+Read [bootstrap-and-marathon-handoff.md](references/bootstrap-and-marathon-handoff.md) when starting a project or handing off into marathon mode.
+
+Bootstrap rules:
+
+- create `codex-continue-novel.ps1` in the project root whenever a new project is initialized
+- build the file from `assets/codex-continue-novel.ps1`
+- replace only the project root placeholder with the actual project root
+- repair stale scripts by rebuilding from the template instead of creating a second variant
+- verify the root script exists and the placeholder is gone before marathon handoff
+- if automatic file creation is blocked, tell the user that startup could not finish automatically and instruct them to create or copy the script manually before marathon handoff
+- do not assume the user will remember the launch command later; record the exact command during handoff
+
 ## Secondary Control View

 When cast density, plotline interference, or foreshadow volume makes linear rereading too blunt, load [graph-and-recall-control.md](references/graph-and-recall-control.md).
@@ -85,6 +109,7 @@ At project launch, major revision, and high-level review, read:
 - [foundational-literary-principles.md](references/foundational-literary-principles.md)
 - [critical-evaluation-standards.md](references/critical-evaluation-standards.md)
 - [epoch-and-people-resonance.md](references/epoch-and-people-resonance.md)
+- [reader-retention-and-ai-failure-modes.md](references/reader-retention-and-ai-failure-modes.md)

 Use them to convert abstract craft into project truth inside:

@@ -132,6 +157,7 @@ Load these references only when their stage is active:
 - [dialogue-writing-rules.md](references/dialogue-writing-rules.md)
 - [suspense-and-reveal-design.md](references/suspense-and-reveal-design.md)
 - [chapter-architecture-rules.md](references/chapter-architecture-rules.md)
+- [chapter-title-method.md](references/chapter-title-method.md)
 - [scene-execution-patterns.md](references/scene-execution-patterns.md)
 - [forgotten-elements-and-line-heat.md](references/forgotten-elements-and-line-heat.md)
 - [authenticity-and-de-ai-pass.md](references/authenticity-and-de-ai-pass.md)
@@ -147,10 +173,12 @@ Stage routing:
 - outline and roadmap design:
  - `graph-and-recall-control.md`
  - `chapter-architecture-rules.md`
+  - `chapter-title-method.md`
  - `character-construction-methods.md`
  - `scene-execution-patterns.md`
 - chapter drafting and revision:
  - always load `chapter-architecture-rules.md`
+  - load `chapter-title-method.md` when the project uses titled chapters or title finalization is active
  - load `graph-and-recall-control.md` when recurrence density, cast rotation, or interference pressure is high
  - load `dialogue-writing-rules.md` when dialogue is carrying pressure
  - load `suspense-and-reveal-design.md` when concealment or reveal fairness is active
@@ -176,7 +204,7 @@ Core questions first:
 4. Ending Direction
   - emotional destination, cost, likely shape of the ending
 5. Style And Market Mode
-   - primary style, support style, target mode, tolerance for slow-burn, forbidden habits
+   - primary style, support style, target mode, tolerance for slow-burn, chapter title mode, forbidden habits

 Language default:

@@ -276,6 +304,19 @@ If the user chooses `draft directly`:
 - flag it as temporary
 - record it in `08-dynamic-state.md` under pending confirmation

+## Chapter Title Control
+
+Read [chapter-title-method.md](references/chapter-title-method.md) when the project uses chapter titles, when the user asks for named chapters, or when a numbered-only chapter system may need to change.
+
+Rules:
+
+- chapter titles are optional; plain numbering is valid when speed, invisibility, or relentless forward pull serves the book better
+- if the project uses chapter titles, lock one naming system in `00-project-overview.md`, `07-chapter-roadmap.md`, and `09-style-guide.md`
+- each title should carry one primary job and at most one secondary job: hook, focus, orientation, motif return, or voice signal
+- generate `3-5` candidate titles from the chapter control card, then choose a working title before drafting
+- do not use chapter titles as blunt summaries, spoiler labels, fake-poetic fog, or generic serial filler
+- after the chapter passes structure and authenticity checks, run a final title-fit recheck and replace the working title if the chapter's true center moved
+
 ## Chapter Workflow

 For every chapter, use this order:
@@ -298,6 +339,7 @@ For every chapter, use this order:
   - drill into deeper style documents only if the chapter needs them
 5. Read only the execution-method references needed for this chapter.
   - always read `chapter-architecture-rules.md`
+   - read `chapter-title-method.md` if the project uses chapter titles or if title finalization is active
   - read `graph-and-recall-control.md` if a retrieval slice was needed
   - read `dialogue-writing-rules.md` if dialogue pressure is central
   - read `suspense-and-reveal-design.md` if suspense or reveal work is active
@@ -319,45 +361,32 @@ For every chapter, use this order:
   - trope convenience overriding human truth
   - lost social or era pressure
 7. Generate a chapter control card using [chapter-control-card.md](references/chapter-control-card.md).
-8. If the risk scan is serious, use the missing information branch.
-9. Draft the chapter from the control card.
+8. If the project uses chapter titles, generate `3-5` candidate titles using [chapter-title-method.md](references/chapter-title-method.md).
+   - choose a working title that matches the project's naming system
+   - record the working title in the chapter control card and `07-chapter-roadmap.md`
+9. If the risk scan is serious, use the missing information branch.
+10. Draft the chapter from the control card.
   - when the chapter needs tighter control, write scene by scene or pressure unit by pressure unit using [scene-execution-patterns.md](references/scene-execution-patterns.md)
-10. Run the chapter benchmark check.
-11. If the benchmark check fails, apply rewrite escalation before accepting the chapter.
-12. Run the authenticity pass using [authenticity-and-de-ai-pass.md](references/authenticity-and-de-ai-pass.md).
-13. Run a post-authenticity mini recheck.
+11. Run the chapter benchmark check.
+12. If the benchmark check fails, apply rewrite escalation before accepting the chapter.
+13. Run the authenticity pass using [authenticity-and-de-ai-pass.md](references/authenticity-and-de-ai-pass.md).
+14. Run a post-authenticity mini recheck.
   - confirm continuity facts still hold
   - confirm character voice and relationship pressure did not flatten or drift
   - confirm hook, closure, and residue still function
-14. Review the chapter for continuity, style integrity, thematic pressure, critical standards, and whether return-pressure handling stayed causal rather than token.
-15. Update dynamic and structural files.
-16. Record the chapter and file updates in the writing log.
+15. If the project uses chapter titles, run the final title check.
+   - confirm the title still matches the accepted chapter's mission, turn, residue, and voice
+   - replace the working title if the chapter changed its center during drafting
+16. Review the chapter for continuity, style integrity, thematic pressure, critical standards, and whether return-pressure handling stayed causal rather than token.
+17. Update dynamic and structural files.
+18. Record the chapter and file updates in the writing log.

 ## Chapter Benchmark Check
-
-After every chapter draft, check at minimum:
-
- character dimensionality increased rather than flattened
- at least one arc moved
- pacing contains hook, propulsion, local closure, and residue appropriate to the benchmark mode
- the chapter closes one obligation and carries at least one debt forward
- theme lands in action, relation, or consequence
- the prose keeps memory points and does not collapse into flat procedural narration
- originality remains intact and the chapter is not benchmark cosplay
- recurring elements either advance, echo, or are intentionally deferred with justified pressure
- scene progression contains at least one changed pressure state through a turn, aftershock, or new obligation
-
-For slow-burn, restrained, anti-formula, or experimental requests:
-
- keep character, logic, theme, and originality as hard gates
- dynamically down-weight hook density, twist frequency, stimulation intensity, and cliff severity
-
-Also check:
-
- protagonist core personality appears in action or dialogue
- dialogue carries pressure rather than only explanation
- suspense or reveal tasks are handled fairly where active
- chapter architecture includes a turn, local closure, and carryover debt
+Read [quality-and-writeback-checks.md](references/quality-and-writeback-checks.md) and [reader-retention-and-ai-failure-modes.md](references/reader-retention-and-ai-failure-modes.md) before accepting a chapter.
+Hard gates:
+- character, logic, theme, and originality stay hard gates even in slow-burn or restrained modes
+- reject flat procedural prose, benchmark cosplay, unsupported payoffs, and dialogue without pressure
+- keep protagonist voice, local closure, and carryover debt legible in the accepted chapter

 ## Forgotten Element Control

@@ -409,6 +438,21 @@ If the chapter benchmark check fails:

 Marathon mode begins only after the user approves the current outline and cast dossier.

+If the user asks for crazy writing, nonstop continuation, auto continuation, or marathon-style hands-off drafting:
+
+- load [bootstrap-and-marathon-handoff.md](references/bootstrap-and-marathon-handoff.md)
+- ensure `codex-continue-novel.ps1` exists in the project root before handoff
+- create or repair the script from `assets/codex-continue-novel.ps1` first if it is missing or stale
+- tell the user to close the current session
+- then tell the user to run this command from the project root:
+
+```powershell
+powershell -ExecutionPolicy Bypass -File .\codex-continue-novel.ps1
+```
+
+- tell the user that `Ctrl+C` stops the looping runner
+- if automatic startup is blocked for any reason, explicitly present the same command as the manual fallback instead of describing the process vaguely
+
 In marathon mode:

 - do not ask the user again chapter by chapter
@@ -419,6 +463,7 @@ For each chapter, still do:
 - required file reads
 - retrieval slice preparation when density requires it
 - chapter control card
+- working title generation and final title recheck when titled chapters are active
 - selected style module loading
 - forgotten-element and line-heat scan
 - benchmark and continuity checks
@@ -441,40 +486,13 @@ Stop marathon mode only when the approved outline has naturally concluded:
 Do not stop because a fixed chapter count or word target was reached.

 ## Dynamic Update Rules
-
-After every chapter, update `08-dynamic-state.md` with:
-
- key events
- character state changes
- relationship changes
- plotline progress
- foreshadow updates
- world or rule changes
- emotional debts, promises, wounds
- retrieval and return pressure
- line heat or cold status where relevant
- last meaningful touch for recurring elements when that affects future recall
- temporary assumptions pending confirmation
- carryover into the next chapter
-
-Use [dynamic-state-template.md](references/dynamic-state-template.md) as the base format.
-
-Update these files when needed:
-
- `03-cast-bible.md` for material character change
- `04-relationship-map.md` for relationship phase change
- `05-main-plotlines.md` for line progression or rerouting
- `06-foreshadow-ledger.md` for planting, activation, or payoff
- `07-chapter-roadmap.md` for future chapter obligations
- `02-worldbuilding.md` for new or corrected world rules
-
-If using a secondary control view:
-
- update canonical files first
- then sync any optional recall or scratch control notes
- never leave the graph view more current than the project files
-
-Apply the writing log rules in [logging-rules.md](references/logging-rules.md).
+Read [quality-and-writeback-checks.md](references/quality-and-writeback-checks.md) for the full writeback checklist.
+After every accepted chapter:
+- update `08-dynamic-state.md` using [dynamic-state-template.md](references/dynamic-state-template.md)
+- update changed canonical files in `03/04/05/06/07/02` as needed
+- if chapter titles are active, make sure the locked chapter title in `07-chapter-roadmap.md` still matches the accepted chapter
+- sync optional secondary control notes only after canonical files are current
+- apply the logging rules in [logging-rules.md](references/logging-rules.md)

 ## Style Module Loading

@@ -503,72 +521,6 @@ Available internal modules:
 - fantasy
 - literary

-## Adaptive Weighting
+## Review And Writeback Detail

-Benchmark standards must adapt to user intent without abandoning quality.
-
-If the user explicitly asks for:
-
- slow-burn
- anti-formula
- restrained intensity
- literary or experimental movement
-
-Then:
-
- keep hard gates on character dimensionality, logical coherence, theme landing, and originality
- down-weight but do not erase hook density, twist frequency, stimulation intensity, and cliff severity
- record the down-weighted checks in `09-style-guide.md`
-
-## Originality And Compliance
-
-Benchmark logic is a pressure system, not a copying license.
-
-Never:
-
- transplant a sample work's core trick
- imitate a sample work's signature cast shell
- preserve a sample work's reveal order with cosmetic changes
- use benchmark labels as an excuse for collage writing
-
-Always:
-
- extract the principle
- restate it as a project-specific action
- test it against the user's actual aims
- prefer fresh combinations of pressure, scene logic, and language
-
-## Quality Gates
-
-Apply [critical-evaluation-standards.md](references/critical-evaluation-standards.md) to chapter review and [epoch-and-people-resonance.md](references/epoch-and-people-resonance.md) to project-scale resonance checks.
-Apply the benchmark layer when the work aims at mainstream readability, market force, or dual-high balance.
-
-Reject or revise a chapter when any of these are true:
-
- a character acts only to push plot or satisfy trope expectation
- a relationship changes without an emotional bridge
- a payoff appears without prior load-bearing setup
- a setting rule changes without document repair
- point of view slips without purpose
- structure is complex on the surface but thin underneath
- an image or motif appears as decoration rather than pressure
- the chapter drops the story's social, institutional, or era force
- benchmark mode requires a hook, closure, or carryover debt and the chapter leaves none
- a benchmark rule group is loaded but not visible in scene logic
- the chapter imitates a sample's recognizable move instead of applying its principle
- prose sounds inflated, fake-deep, dead, or slogan-like
- a nonlinear shift reveals nothing about character truth or theme
- style modules are mixed without purpose
-
-## Common Mistakes
-
- Treating project files as optional
- Building only one visible line and calling it a long-form structure
- Writing a chapter before deciding what it must advance
- Forgetting to update dynamic state after drafting
- Recording key truths in the log instead of the real project files
- Treating foundational literary principles as inspiration instead of constraints
- Loading every style module instead of only the selected internal modules
- Treating a scratch graph or recall map as if it were the canon
- Bringing back cold elements with empty cameos instead of causal pressure
- Flattening prose while trying to remove AI patterns
+Read [quality-and-writeback-checks.md](references/quality-and-writeback-checks.md) when applying adaptive weighting, originality discipline, chapter quality gates, final writeback, or common-failure review.
@@ -0,0 +1,57 @@
+# codex-continue-novel.ps1
+# powershell -ExecutionPolicy Bypass -File .\codex-continue-novel.ps1
+
+$followUp = "使用novel-control-station技能继续小说的疯狂创作。如果整本小说已经完成，而不是当前章节完成，请直接回复AllNovelDone"
+$logFile = ".\codex-continue.log"
+$projectRoot = "__PROJECT_ROOT__"
+
+function Log($msg) {
+    $time = Get-Date -Format "yyyy-MM-dd HH:mm:ss"
+    $line = "[$time] $msg"
+    Write-Host $line
+    Add-Content -Path $logFile -Value $line
+}
+
+function Write-CommandOutput($lines) {
+    foreach ($line in @($lines)) {
+        if ($null -eq $line) {
+            continue
+        }
+
+        $text = [string]$line
+        Write-Host $text
+        Add-Content -Path $logFile -Value $text
+    }
+}
+
+function Test-AllNovelDone($lines) {
+    foreach ($line in @($lines)) {
+        if ([string]$line -match '^\s*AllNovelDone\s*$') {
+            return $true
+        }
+    }
+
+    return $false
+}
+
+Log "开始恢复最后一次会话"
+
+while ($true) {
+    Set-Location -LiteralPath $projectRoot
+    $output = & codex exec resume --last $followUp --skip-git-repo-check 2>&1
+    $code = $LASTEXITCODE
+    Write-CommandOutput $output
+
+    if (Test-AllNovelDone $output) {
+        Log "收到 AllNovelDone，整本小说已完成，停止脚本"
+        break
+    }
+
+    if ($code -eq 0) {
+        Log "会话正常结束"
+        continue
+    }
+
+    Log "检测到异常退出，退出码: $code，准备自动继续"
+    Start-Sleep -Seconds 2
+}
@@ -1,93 +0,0 @@
-# Baseline Scenarios
-
-These scenarios define what the fiction system should prevent or improve.
-
-## Scenario 1: New Project Alignment
-
- Prompt shape: user gives a rough premise and asks to start the novel
- Baseline failure without skill: the agent starts outlining or drafting immediately
- Desired behavior: ask alignment questions first, then create standard files
-
-## Scenario 2: Continuation With Existing Files
-
- Prompt shape: user asks for the next chapter after prior chapters and project files exist
- Baseline failure without skill: the agent drafts from memory and misses obligations
- Desired behavior: read core files, generate chapter control card, then draft
-
-## Scenario 3: Contradiction And Drift
-
- Prompt shape: user continues after conflicting notes, missing updates, and relationship drift
- Baseline failure without skill: the agent ignores conflict and compounds the drift
- Desired behavior: identify risk, branch to refine details or draft directly, then update dynamic state
-
-## Scenario 4: Foundational Canon
-
- Prompt shape: user asks for a high-end long novel and expects literary depth, not trope stacking
- Baseline failure without skill: the agent offers market formula, fake depth, and ungrounded prestige language
- Desired behavior: load the foundational canon, translate it into project files, and enforce causal, human, linguistic, and epoch-level standards before drafting
-
-## Scenario 5: Benchmark Outline
-
- Prompt shape: user asks for a hot-and-reputable novel outline, but the initial outline is slogan-heavy and character-light
- Baseline failure without skill: the agent accepts the hollow outline and moves into prose
- Desired behavior: run outline benchmark checks, reject the weak outline, and repair it before drafting
-
-## Scenario 6: Benchmark Chapter Rewrite
-
- Prompt shape: user asks for the next chapter and the first draft has weak hook, flat arc movement, and no meaningful residue
- Baseline failure without skill: the agent accepts the draft without rewrite
- Desired behavior: run chapter benchmark checks, force rewrites, and escalate correctly on repeated failure
-
-## Scenario 7: Adaptive Weighting
-
- Prompt shape: user asks for a restrained, slow-burn, literary chapter with clear anti-formula intent
- Baseline failure without skill: the agent imposes mass-market hook density and damages the requested tone
- Desired behavior: keep hard quality gates while down-weighting pace-intensity metrics
-
-## Scenario 8: Originality And Compliance
-
- Prompt shape: user asks for a chapter inspired by famous recent hits
- Baseline failure without skill: the agent imitates signature twists, cast shells, or recognizable setups
- Desired behavior: extract principles, not shells, and preserve originality
-
-## Scenario 9: Internal Style Modules
-
- Prompt shape: user asks for a chapter with selected styles and expects style loading to stay selective
- Baseline failure without skill: the agent still depends on old external style skills or loads all modules at once
- Desired behavior: route style loading through internal module docs inside `novel-control-station` and keep progressive disclosure
-
-## Scenario 10: Interview Derivation
-
- Prompt shape: user gives a genre, protagonist core personality, and long-form ambition that should trigger follow-up questions
- Baseline failure without skill: the agent asks only a fixed shallow questionnaire
- Desired behavior: ask core questions first, then derive more questions from the answers
-
-## Scenario 11: Outline And Cast Handoff
-
- Prompt shape: user completes the interview and asks to start the novel
- Baseline failure without skill: the agent starts drafting without presenting full outline and cast dossier
- Desired behavior: present full outline and cast dossier before any chapter drafting
-
-## Scenario 12: Direct-Edit Branch
-
- Prompt shape: user asks to adjust the current outline and cast after seeing them
- Baseline failure without skill: the agent restarts the whole interview or rebuilds from scratch
- Desired behavior: revise the existing plan directly and show the updated version
-
-## Scenario 13: Marathon Mode
-
- Prompt shape: user approves outline and cast, then asks for nonstop continuous creation
- Baseline failure without skill: the agent treats nonstop mode as no-check mode
- Desired behavior: keep chapter control, benchmark checks, document updates, and logging while continuing automatically
-
-## Scenario 14: Natural Finish
-
- Prompt shape: user wants the whole novel completed in marathon mode
- Baseline failure without skill: the agent stops by arbitrary chapter count or word target
- Desired behavior: continue until the approved outline naturally concludes
-
-## Scenario 15: Dialogue Character Suspense Structure
-
- Prompt shape: user requests a chapter where character, dialogue, suspense, and structure all matter
- Baseline failure without skill: the agent ignores the new execution-method references and writes by generic instinct
- Desired behavior: apply the new reference layer and make those pressures visible in the chapter plan and review
@@ -1,24 +0,0 @@
-# Scenario 01: New Project Alignment
-
-## Prompt
-
-"I want a long serialized novel about a disgraced heir returning to a port city controlled by guilds and hidden courts. Start planning and write chapter one."
-
-## Expected Baseline Failure Without The Skill
-
- jumps into plotting or drafting immediately
- does not ask for audience, tone, theme, line structure, or ending direction
- treats the novel like a one-shot prompt
-
-## Required Skill Behavior
-
- ask alignment questions before formal design
- clarify genre, audience, cast pressure, plot direction, thematic question, ending direction, and style choices
- refuse to begin full design until main plot direction, core character tension, and ending direction are aligned
- create the standard document set before drafting chapter one
-
-## Pass Criteria
-
- no chapter drafting before alignment
- standard files are explicitly established
- the structure is multi-line by default
@@ -1,27 +0,0 @@
-# Scenario 02: Continuation With Existing Files
-
-## Prompt
-
-"Continue chapter 27. Use the current files. The heroine must push the smuggling line forward, the rival line cannot disappear, and the confession debt from chapter 24 still matters."
-
-## Expected Baseline Failure Without The Skill
-
- drafts from loose memory
- ignores at least one active line
- forgets prior emotional debt
- uses style loosely without controlled loading
-
-## Required Skill Behavior
-
- read the required project files before drafting
- read only the relevant style modules for the chapter
- generate a chapter control card
- identify line obligations, emotional debt, style goals, and risks
- draft from the control card rather than improvising freely
-
-## Pass Criteria
-
- chapter control card exists before drafting
- multiple lines remain active
- prior emotional debt affects the chapter
- style loading is selective rather than global
@@ -1,28 +0,0 @@
-# Scenario 03: Contradiction And Drift
-
-## Prompt
-
-"Write the next chapter. The notes say the mentor died in chapter 9, but the outline for chapter 18 has him speaking in person. Also the relationship map still says the siblings are estranged, but chapter 17 ended with a fragile reunion."
-
-## Expected Baseline Failure Without The Skill
-
- drafts over the contradiction
- preserves outdated state
- introduces more inconsistency
- fails to document repair decisions
-
-## Required Skill Behavior
-
- identify every conflicting or risky item
- explain what each conflict threatens
- ask whether to refine details or draft directly
- if drafting directly, use only minimal temporary assumptions
- update `08-dynamic-state.md` and any impacted structural files after drafting
- append an audit entry to the writing log
-
-## Pass Criteria
-
- risk branch occurs before drafting
- contradiction handling is explicit
- state repair is written back to project documents
- the writing log remains audit-only
@@ -1,30 +0,0 @@
-# Scenario 04: Foundational Canon
-
-## Prompt
-
-"Design a major long novel about a family empire built on an industrial river, a daughter returning after a cover-up, and a city whose prosperity depends on what everyone agrees not to see. I want world-class literary depth, but it still has to read clearly."
-
-## Expected Baseline Failure Without The Skill
-
- defaults to genre tropes and prestige-like language
- speaks about "depth" without defining human pressure
- ignores institutions, labor, class, public force, and ordinary life
- treats structure and point of view as decoration
- does not define ending residue, causal pressure, or language constraints
-
-## Required Skill Behavior
-
- read the foundational canon before project design or major revision
- convert those principles into standard project files rather than leaving them abstract
- define character desire, fear, shame, debt, contradiction, and social position
- define causal line pressure, social or era pressure, and ending direction
- define language discipline, image obligations, and formal limits
- reject fake depth, event pile, decorative motifs, and ornamental nonlinear structure during review
-
-## Pass Criteria
-
- `01-theme-and-proposition.md` records a real central question, opposing values, and human cost
- `03-cast-bible.md` records pressure-rich characters rather than role labels
- `05-main-plotlines.md` records causal chains and social or era force
- `09-style-guide.md` records language standards and forbidden drift
- the design remains readable on the surface while carrying deeper structural pressure
@@ -1,26 +0,0 @@
-# Scenario 05: Benchmark Outline
-
-## Prompt
-
-"Give me a dual-high outline for a historical family saga with political struggle and a female doctor line. I want it to be deep, popular, and emotionally addictive."
-
-## Expected Baseline Failure Without The Skill
-
- accepts a slogan-heavy outline
- gives role-label characters with no pressure system
- keeps plotlines parallel instead of interfering
- treats the ending as a vague mood instead of a direction with closure and residue
-
-## Required Skill Behavior
-
- load the benchmark common laws and the relevant genre rules
- run an outline benchmark check before allowing prose drafting
- reject hollow theme statements and pressure-light character design
- repair line interference, ending direction, and genre promise before chapter drafting
-
-## Pass Criteria
-
- the outline is blocked until benchmark issues are repaired
- character desire, obstacle, price, and arc direction are explicit
- multiple lines interfere materially
- the benchmark group is visible in the outline logic
@@ -1,25 +0,0 @@
-# Scenario 06: Benchmark Chapter Rewrite
-
-## Prompt
-
-"Write chapter 18. Keep the romance warm, the political line active, and the investigation line from going cold."
-
-## Expected Baseline Failure Without The Skill
-
- drafts one weak version and accepts it
- gives the chapter movement but no strong hook, no local closure, and no carryover debt
- leaves character arc movement vague
-
-## Required Skill Behavior
-
- run a chapter benchmark check after the first draft
- force a full rewrite on the first failure
- force a targeted rewrite on the second failure
- on the third failure, stop blind rewriting and summarize cause-directed fixes
-
-## Pass Criteria
-
- the rewrite escalation follows the configured 1/2/3 sequence
- the accepted chapter shows hook, propulsion, closure, and residue
- at least one character arc increment is explicit
- the result is logged in audit form
@@ -1,25 +0,0 @@
-# Scenario 07: Adaptive Weighting
-
-## Prompt
-
-"Design a slow-burn literary mystery. I want pressure and depth, but I do not want constant twists, loud cliffhangers, or exaggerated commercial pacing."
-
-## Expected Baseline Failure Without The Skill
-
- mistakes slow-burn for weakness or emptiness
- overcorrects by adding forced hooks and high-frequency reversals
- damages the requested restrained tone
-
-## Required Skill Behavior
-
- keep character depth, logic, theme landing, and originality as hard gates
- down-weight hook density, twist frequency, stimulation intensity, and cliff severity
- record the down-weighted checks in the project style state
- preserve clear reader orientation and residue
-
-## Pass Criteria
-
- the tone remains restrained without becoming flat
- hard quality gates still apply
- pace-intensity metrics are visibly down-weighted rather than removed
- the benchmark layer adapts instead of forcing formula
@@ -1,24 +0,0 @@
-# Scenario 08: Originality And Compliance
-
-## Prompt
-
-"Make it feel like the smartest parts of recent historical hits and suspense hits, but fresher and more literary."
-
-## Expected Baseline Failure Without The Skill
-
- imitates a recognizable setup, cast shell, twist engine, or voice texture
- confuses benchmarking with collage writing
- keeps only surface novelty
-
-## Required Skill Behavior
-
- extract principles from benchmark samples instead of copying their shells
- restate those principles as project-specific constraints
- reject signature-trick imitation during review
- preserve originality and compliance even under strong benchmark pressure
-
-## Pass Criteria
-
- no recognizable sample-specific shell survives intact
- the draft can explain which principles were extracted and how they were transformed
- originality and compliance are treated as hard gates
@@ -1,25 +0,0 @@
-# Scenario 09: Internal Style Modules
-
-## Prompt
-
-"Continue chapter 22 in a dual-style mode: primary suspense, support romance. Keep the mystery line readable, but do not load irrelevant style systems."
-
-## Expected Baseline Failure Without The Skill
-
- still points to deleted external style skills
- loads all style guidance instead of the selected modules
- loses progressive disclosure after the merge
-
-## Required Skill Behavior
-
- route style loading through `novel-control-station/references/style-modules/index.md`
- read only the selected internal module `core.md` files first
- read deeper style documents only when the chapter requires them
- preserve the limit of 1 or 2 primary styles and at most 1 support style
-
-## Pass Criteria
-
- no deleted external style-skill dependency remains in active skill flow
- internal style modules are referenced explicitly
- style loading remains selective rather than global
- the merged system still writes style pressure into chapter control
@@ -1,27 +0,0 @@
-# Scenario 10: Interview Derivation
-
-## Prompt
-
-"I want a long historical mystery. The protagonist is brilliant but compulsively overcontrols everything, and I want the whole thing to be multi-line and psychologically tense."
-
-## Expected Baseline Failure Without The Skill
-
- asks only generic setup questions
- never derives follow-up questions from the genre, personality, or long-form scope
- leaves clue logic, power chain, stress behavior, and line interference underdefined
-
-## Required Skill Behavior
-
- ask core startup questions first
- derive more questions from the answers
- follow up on:
-  - historical institution and era pressure
-  - mystery clue fairness and reveal structure
-  - protagonist stress behavior and speech signature
-  - multi-line interference and finish logic
-
-## Pass Criteria
-
- the interview branches from the user's answers
- follow-up questions are project-specific rather than generic
- the answers gathered are sufficient to support outline and cast design
@@ -1,24 +0,0 @@
-# Scenario 11: Outline And Cast Handoff
-
-## Prompt
-
-"Good, that's the setup I want. Start chapter one now."
-
-## Expected Baseline Failure Without The Skill
-
- jumps into drafting immediately after interview alignment
- never presents the full outline
- never presents the cast dossier
-
-## Required Skill Behavior
-
- block chapter drafting
- present the full outline first
- present the full cast dossier first
- wait for user approval or revision request before beginning the novel
-
-## Pass Criteria
-
- no prose chapter is drafted before outline and cast handoff
- cast dossier includes personality, contradiction, arc direction, and speech signature
- outline includes lines, stage progression, key turns, and ending direction
@@ -1,24 +0,0 @@
-# Scenario 12: Direct-Edit Branch
-
-## Prompt
-
-"Keep the current story, but change the male lead from cold and restrained to glib and evasive, and move the betrayal reveal one phase later."
-
-## Expected Baseline Failure Without The Skill
-
- restarts the whole interview
- throws away the current outline and cast instead of revising them
- loses already aligned material
-
-## Required Skill Behavior
-
- enter the direct-edit branch
- modify the existing outline and cast dossier in place
- preserve non-conflicting approved material
- show the revised version
-
-## Pass Criteria
-
- the system revises the current plan instead of rebooting
- only the requested areas and their dependencies are changed
- the updated outline and cast remain coherent
@@ -1,23 +0,0 @@
-# Scenario 13: Marathon Mode
-
-## Prompt
-
-"The outline and character files look right. Turn on marathon mode and write straight through the book without asking me chapter by chapter."
-
-## Expected Baseline Failure Without The Skill
-
- treats nonstop writing as permission to skip control logic
- stops updating dynamic state or structural files
- stops logging benchmark failures and rewrites
-
-## Required Skill Behavior
-
- enter marathon mode only after approved outline and cast handoff
- keep every per-chapter control step active
- continue automatically only after each chapter is accepted and written back into project files
-
-## Pass Criteria
-
- marathon mode does not disable chapter control
- dynamic state, structural files, and writing log still update every chapter
- automatic continuation happens only after internal acceptance
@@ -1,24 +0,0 @@
-# Scenario 14: Natural Finish
-
-## Prompt
-
-"Stay in marathon mode until the whole novel is done."
-
-## Expected Baseline Failure Without The Skill
-
- uses a fixed chapter count or word target as the stop rule
- rushes line resolution to force an ending
- stops before major debts and support lines are resolved
-
-## Required Skill Behavior
-
- stop only when the approved outline naturally concludes
- resolve main lines and major debts
- close or deliberately transform support lines with justified residue
- fulfill the established ending direction
-
-## Pass Criteria
-
- stop condition is story completion, not arbitrary size
- major lines and debts are tracked to closure
- ending direction is fulfilled rather than abruptly truncated
@@ -1,25 +0,0 @@
-# Scenario 15: Dialogue Character Suspense Structure
-
-## Prompt
-
-"Write a chapter where the heroine's controlling personality shows through the dialogue, a hidden clue changes the power balance, and the chapter should end on a necessary next move rather than a random cliff."
-
-## Expected Baseline Failure Without The Skill
-
- writes generic dialogue
- treats character as surface label
- uses suspense as withholding without fair pressure
- produces a chapter with events but weak architecture
-
-## Required Skill Behavior
-
- apply the new execution-method references during planning and review
- make core personality visible in behavior and speech
- make suspense fair, pressured, and consequential
- give the chapter an entry, turn, local closure, and carryover debt
-
-## Pass Criteria
-
- chapter planning visibly uses character, dialogue, suspense, and structure rules
- the resulting chapter would pass those execution-method checks
- the ending creates necessity rather than empty noise
@@ -0,0 +1,47 @@
+# Bootstrap And Marathon Handoff
+
+Use this reference only when:
+
+- starting a new novel project
+- preparing crazy-writing or auto-continuation handoff
+
+## Startup checklist
+
+1. Create the standard project files first.
+2. Read `assets/codex-continue-novel.ps1`.
+3. Replace `__PROJECT_ROOT__` with the actual absolute project root.
+4. Write the result to `codex-continue-novel.ps1` in the project root.
+5. Verify that `codex-continue-novel.ps1` exists in the project root.
+6. Verify that the written script no longer contains `__PROJECT_ROOT__`.
+7. Record the launch command in the handoff wording:
+
+```powershell
+powershell -ExecutionPolicy Bypass -File .\codex-continue-novel.ps1
+```
+
+## Handoff checklist
+
+1. Confirm the outline and cast dossier are already approved.
+2. Confirm the root runner script exists and is current.
+3. Tell the user to close the current session.
+4. Tell the user to run:
+
+```powershell
+powershell -ExecutionPolicy Bypass -File .\codex-continue-novel.ps1
+```
+
+5. Tell the user that `Ctrl+C` stops the looping runner.
+6. If automatic bootstrap or automatic startup was blocked, say so plainly and present the same command as the manual fallback.
+
+## Example
+
+Example startup handoff:
+
+- project root: `D:\novels\qing-ye-ji`
+- action: write `D:\novels\qing-ye-ji\codex-continue-novel.ps1` from the template, replacing only `__PROJECT_ROOT__`
+- user-facing handoff: `关闭当前会话后，在项目根目录运行 powershell -ExecutionPolicy Bypass -File .\codex-continue-novel.ps1`
+
+Example marathon handoff:
+
+- user request: `开启疯狂写作`
+- required response: confirm the runner script exists, tell the user to close the current session, then present the exact command block without paraphrasing it away
@@ -78,13 +78,22 @@ Not every chapter should use the same structure. Rotate among:

 Variation helps long fiction stay alive.

-## 7. Structure Must Match Mode
+## 7. Chapter Title Must Match The Chapter Itself
+
+When the project uses chapter titles:
+
+- the title should reinforce mission, turn, residue, orientation, or motif
+- the title should not summarize the chapter's full event chain
+- the title should create pull without stealing the chapter's discovery
+- the title may be revised after drafting if the chapter's real center changed
+
+## 8. Structure Must Match Mode

 - heat-forward mode may push sharper hooks and tighter chapter-end pull
 - reputation-forward mode may lean harder on residue and thematic pressure
 - dual-high balance needs both readability and aftereffect

-## 8. Chapter Revision Questions
+## 9. Chapter Revision Questions

 - where does this chapter truly begin
 - what is the pivot
@@ -93,7 +102,7 @@ Variation helps long fiction stay alive.
 - what line went cold and needs attention
 - what human pressure became clearer here

-## 9. Failure Signals
+## 10. Failure Signals

 - chapter exists only to move people into place
 - start is too early
@@ -17,6 +17,17 @@ Generate this before drafting every chapter.
 - required_plotlines:
 - required_prior_debts:

+## Chapter Title
+- title_mode:
+- title_system_reference:
+- primary_title_job:
+- secondary_title_job:
+- title_anchor:
+- title_candidates:
+- working_title:
+- spoiler_risk:
+- final_title_recheck:
+
 ## Recall And Return Pressure
 - retrieval_slice_targets:
 - cold_lines_due:
@@ -108,4 +119,7 @@ Minimum standard:
 - every chapter must state how benchmark expectations were met or intentionally down-weighted
 - every chapter must define dialogue, suspense, or structure tasks when those pressures are active
 - every chapter must mark authenticity focus when prose risks generic drift
+- if titled chapters are active, every control card must generate `3-5` candidates and choose one working title
+- a chapter title must signal pressure, orientation, or residue rather than summarize the whole chapter
+- a working title may change after drafting if the chapter's true center moved
 - every hook must connect to an existing line, not a random surprise
@@ -0,0 +1,140 @@
+# Chapter Title Method
+
+Use this when the project uses chapter titles, when the user asks to name or rename chapters, or when a numbered-only chapter system may need to change.
+
+Goal: make chapter titles act as controlled story signals instead of decorative summaries.
+
+## 1. First Decide Whether This Project Needs Chapter Titles
+
+Chapter titles are optional.
+
+Use numbered chapters only when that better serves:
+
+- invisibility
+- speed
+- relentless forward pull
+- a deliberately plain presentation
+
+Use titled chapters when titles materially add:
+
+- hook value
+- point-of-view or time/place orientation
+- motif return
+- tonal identity
+- structural contrast between chapters
+
+If a title adds nothing real, keep the chapter numbered.
+
+## 2. Lock One Naming System For The Whole Book
+
+Before naming individual chapters, decide the book-level system and record it in `00-project-overview.md` and `09-style-guide.md`.
+
+Common systems:
+
+- pure numbers
+- number plus short title
+- character or point-of-view labels
+- time or place anchors
+- image or motif titles
+- pressure phrases
+- object, clue, ritual, promise, or line-of-dialogue titles
+
+Do not mix systems casually. A format shift is allowed only when it marks a real structural phase change and is written into project canon.
+
+## 3. Give Each Title A Job
+
+Each chapter title should carry one primary job and at most one secondary job:
+
+- focus the chapter's core pressure or change
+- create curiosity without stealing discovery
+- orient the reader to point of view, time, or place
+- return a motif, image, or theme-bearing object
+- signal a special chapter mode or narrative stance
+
+If a title is trying to do everything, it usually becomes weak.
+
+## 4. Build The Title From The Chapter Control Card
+
+Pull possible title anchors from:
+
+- core goal
+- key conflict
+- dominant residue
+- point of view
+- time or place marker
+- returning image or motif
+- object, promise, lie, debt, or clue under pressure
+- chapter-end hook
+
+Choose the most load-bearing anchor, not the chapter's entire event list.
+
+## 5. Generate 3 To 5 Candidates
+
+Before locking the title, generate multiple options from different angles:
+
+- pressure phrase
+- concrete image or object
+- contradiction
+- charged verb plus noun
+- speaker or point-of-view marker
+- time or place anchor
+
+Prefer precise nouns, active verbs, and language that sounds native to this book's voice.
+
+## 6. Screen Candidates With Five Checks
+
+Pick the best candidate by checking:
+
+- chapter fit: does it still match the actual chapter center
+- curiosity: does it create pull without giving away the payoff
+- distinctiveness: does it avoid blending into nearby chapter titles
+- voice: does it sound like this book, genre, narrator, and market mode
+- consistency: does it match the project's naming system and title register
+
+Reject titles that are spoilers, flat summaries, vague decorative riddles, or recycled grammatical shells.
+
+## 7. Timing Rules
+
+Use a two-stage process:
+
+1. before drafting, record `3-5` candidates and choose a working title in the chapter control card
+2. after the chapter passes structure and authenticity checks, run a final title-fit recheck and lock the final title
+
+If the chapter's true center moved during drafting, change the title instead of defending the first idea.
+
+## 8. Genre-Pressure Adjustments
+
+- suspense and thriller chapters often benefit from pressure, threat, false certainty, or charged-object titles
+- mystery chapters often benefit from clue, witness, question, or location framing
+- romance chapters often benefit from emotional debt, misread promise, or relational shift framing
+- literary chapters often benefit from image, contradiction, or motif return
+- historical and fantasy chapters often benefit from institution, place, ritual, or cost framing
+
+These are pressure preferences, not templates.
+
+## 9. Anti-Patterns
+
+- title summarizes the chapter's climax or reveal
+- title sounds dramatic but the chapter does not support it
+- title is generic enough to fit twenty other chapters
+- title system changes every few chapters with no structural reason
+- title carries information the chapter itself never earns
+- title is frozen before drafting and never rechecked after the chapter changes
+
+## 10. Default Recommendation
+
+When no special project constraint says otherwise:
+
+- use chapter titles only if they add real value
+- keep one stable naming system for the whole book
+- let each title do one main job
+- generate several candidates from the control card
+- finalize the title only after the accepted chapter exists
+
+## 11. 中文执行模板
+
+在需要直接执行章节起名时，可直接使用下面这段模板：
+
+```text
+如果这本书尚未确定是否使用章节名，先判断章节名是否真的有增益；若章节名不能提升钩子、聚焦、定位、意象回收或叙述声音，就保留纯编号。若本书使用章节名，先读取 00-project-overview.md、07-chapter-roadmap.md、09-style-guide.md 和本章 chapter control card，确认全书命名体系、标题 register、标题主职责。然后从本章的 core goal、key conflict、dominant residue、point of view、motif/object/clue、chapter-end hook 中提取最承重的 1 到 2 个命名锚点，按不同角度生成 3 到 5 个候选标题。每个标题只承担 1 个主任务，最多兼顾 1 个副任务：制造悬念、聚焦压力、提供时间/地点/视角定位、回收意象、突出本书声音。优先使用具体名词、主动动词、带压力的短语；避免剧透高潮、复述流水账、空泛文艺句、伪诗意迷雾、模板化词壳和与全书体系不一致的格式。随后按 5 项标准筛选：1. 贴章度，2. 悬念度，3. 辨识度，4. 声音匹配度，5. 体系一致性；选出一个 working title 写入章节控制卡。正文成稿并通过结构、真实性、连续性检查后，再复核标题是否仍然准确对应本章真正的 mission、turn、residue 和 ending hook；如果本章重心已经变化，立即改题，不要硬保留旧题。最终输出时给出：本书标题体系一句话、本章 3 到 5 个候选标题、每个候选标题的命名理由、最终定稿标题，以及一句不超过 20 字的定稿说明。
+```
@@ -95,15 +95,17 @@ Loop:
 2. prepare retrieval slice if density requires it
 3. run forgotten-element and line-heat scan
 4. generate chapter control card
-5. load active style modules
-6. draft chapter
-7. run benchmark and continuity checks
-8. rewrite if needed
-9. run authenticity pass
-10. run a post-authenticity mini recheck
-11. update dynamic and structural files
-12. log the chapter
-13. continue to the next chapter
+5. generate a working title if titled chapters are active
+6. load active style modules
+7. draft chapter
+8. run benchmark and continuity checks
+9. rewrite if needed
+10. run authenticity pass
+11. run a post-authenticity mini recheck
+12. run the final title recheck if titled chapters are active
+13. update dynamic and structural files
+14. log the chapter
+15. continue to the next chapter

 ## 7. Rewrite Escalation Still Applies

@@ -145,3 +147,4 @@ When the novel approaches conclusion, ask:
 - optional secondary control notes left newer than the canonical files
 - authenticity cleanup flattening character or genre voice
 - authenticity cleanup changing facts or hook pressure without a final recheck
+- titled chapters generated on autopilot with no fit check against the accepted chapter
@@ -18,6 +18,9 @@ If graph-style control is needed, derive it from these files instead of creating
 - target length
 - user approval state
 - target mode: heat, reputation, dual-high balance
+- chapter title mode: numbered, titled, hybrid
+- chapter title system
+- chapter title primary purpose
 - core promise
 - primary hooks
 - historical or social pressure
@@ -117,6 +120,10 @@ For each item:
 For each planned chapter:

 - chapter number or slot
+- working title
+- locked title
+- title primary job
+- title system slot
 - stage target
 - main objective
 - main-line push
@@ -138,6 +145,7 @@ Use the dedicated template in [dynamic-state-template.md](dynamic-state-template
 - selected primary styles
 - support style
 - forbidden styles
+- chapter title register
 - Chinese register mode
 - prose density
 - dialogue ratio
@@ -145,6 +153,7 @@ Use the dedicated template in [dynamic-state-template.md](dynamic-state-template
 - pacing preference
 - benchmark group
 - dynamic down-weighted checks
+- payoff expectations for active primary styles
 - forbidden trope or market habits
 - image system
 - language taboos
@@ -0,0 +1,117 @@
+# Quality And Writeback Checks
+
+Use this reference when:
+
+- accepting or rejecting a drafted chapter
+- applying adaptive weighting for slow-burn or restrained requests
+- running final authenticity and continuity review
+- updating canonical files after an accepted chapter
+
+## Chapter benchmark check
+
+After every chapter draft, confirm at minimum:
+
+- character dimensionality increased rather than flattened
+- at least one arc moved
+- pacing contains hook, propulsion, local closure, and residue appropriate to the active benchmark mode
+- the chapter closes one obligation and carries at least one debt forward
+- theme lands in action, relation, or consequence
+- the prose keeps memory points and does not collapse into flat procedural narration
+- originality remains intact and the chapter is not benchmark cosplay
+- recurring elements either advance, echo, or are intentionally deferred with justified pressure
+- scene progression contains at least one changed pressure state through a turn, aftershock, or new obligation
+- protagonist core personality appears in action or dialogue
+- dialogue carries pressure rather than only explanation
+- suspense or reveal tasks are handled fairly where active
+- chapter architecture includes a turn, local closure, and carryover debt
+- if chapter titles are active, the selected title fits the accepted chapter, preserves curiosity, and matches the project's naming system
+
+## Dynamic update rules
+
+After every accepted chapter, update `08-dynamic-state.md` with:
+
+- key events
+- character state changes
+- relationship changes
+- plotline progress
+- foreshadow updates
+- world or rule changes
+- emotional debts, promises, wounds
+- retrieval and return pressure
+- line heat or cold status where relevant
+- last meaningful touch for recurring elements when that affects future recall
+- temporary assumptions pending confirmation
+- carryover into the next chapter
+
+Then update these files when needed:
+
+- `03-cast-bible.md`
+- `04-relationship-map.md`
+- `05-main-plotlines.md`
+- `06-foreshadow-ledger.md`
+- `07-chapter-roadmap.md`
+- `02-worldbuilding.md`
+
+If a secondary control view exists, update canonical files first and never leave the sidecar newer than canon.
+If chapter titles are active and the accepted title changed during drafting, repair the title fields in `07-chapter-roadmap.md` during the same writeback pass.
+
+## Adaptive weighting
+
+If the user explicitly asks for slow-burn, anti-formula, restrained intensity, or literary / experimental movement:
+
+- keep character, logic, theme, and originality as hard gates
+- down-weight but do not erase hook density, twist frequency, stimulation intensity, and cliff severity
+- record the down-weighted checks in `09-style-guide.md`
+
+## Quality gates
+
+Apply [critical-evaluation-standards.md](critical-evaluation-standards.md) to chapter review and [epoch-and-people-resonance.md](epoch-and-people-resonance.md) to project-scale resonance checks.
+Reject or revise a chapter when any of these are true:
+
+- a character acts only to push plot or satisfy trope expectation
+- a relationship changes without an emotional bridge
+- a payoff appears without prior load-bearing setup
+- a setting rule changes without document repair
+- point of view slips without purpose
+- structure is complex on the surface but thin underneath
+- an image or motif appears as decoration rather than pressure
+- the chapter drops the story's social, institutional, or era force
+- benchmark mode requires a hook, closure, or carryover debt and the chapter leaves none
+- a benchmark rule group is loaded but not visible in scene logic
+- the chapter imitates a sample's recognizable move instead of applying its principle
+- prose sounds inflated, fake-deep, dead, or slogan-like
+- a nonlinear shift reveals nothing about character truth or theme
+- style modules are mixed without purpose
+
+## Originality discipline
+
+Benchmark logic is a pressure system, not a copying license.
+
+Never:
+
+- transplant a sample work's core trick
+- imitate a sample work's signature cast shell
+- preserve a sample work's reveal order with cosmetic changes
+- use benchmark labels as an excuse for collage writing
+
+Always:
+
+- extract the principle
+- restate it as a project-specific action
+- test it against the user's actual aims
+- prefer fresh combinations of pressure, scene logic, and language
+
+## Common mistakes
+
+- treating project files as optional
+- building only one visible line and calling it a long-form structure
+- writing a chapter before deciding what it must advance
+- forgetting to update dynamic state after drafting
+- recording key truths in the log instead of the real project files
+- treating foundational literary principles as inspiration instead of constraints
+- loading every style module instead of only the selected internal modules
+- treating a scratch graph or recall map as if it were the canon
+- bringing back cold elements with empty cameos instead of causal pressure
+- flattening prose while trying to remove AI patterns
+- locking a chapter title before drafting and never rechecking whether the accepted chapter still earns it
+- using spoiler summaries or generic chapter labels when a titled system is active
@@ -0,0 +1,85 @@
+# Reader Retention And AI Failure Modes
+
+Use this reference only for broad, cross-genre risks that damage long-form readability or frequently appear in AI-assisted drafting.
+
+## Universal and necessary filters
+
+Keep these as global checks because they are broadly applicable across genres and formats:
+
+- opening drag
+- info dumping and over-explanation
+- character motivation blur
+- long-range drift
+- repetition and formulaic language
+
+Do not upgrade genre-specific tastes into universal law here. Keep things like niche reveal timing, romance triangle structure, or detective-canon specifics inside genre modules.
+
+## Opening drag
+
+Problem:
+
+- the novel or chapter spends too long clearing its throat
+- nothing meaningful is at risk yet
+- the reader cannot tell why this scene exists now
+
+What to protect:
+
+- open with live pressure, asymmetry, risk, want, or disruption
+- let the chapter mission become legible early
+- keep throat-clearing, weather, backstory, and neutral setup subordinate to present pressure
+
+## Info dumping and over-explanation
+
+Problem:
+
+- the text explains the world, motive, or theme in blocks that halt motion
+- subtext gets flattened into explicit commentary
+- the reader is denied the work of inference
+
+What to protect:
+
+- let explanation ride inside conflict, choice, friction, and consequence
+- prefer concrete evidence over abstract summary
+- trust the reader to connect nearby signals before adding more explanation
+
+## Character motivation blur
+
+Problem:
+
+- the protagonist or core cast acts without a legible want, fear, debt, or contradiction
+- side characters become function carriers instead of people under pressure
+- emotional turns happen before their bridge is visible
+
+What to protect:
+
+- keep motive, stress response, and contradiction readable in action
+- make important turns feel caused, not assigned
+- do not let convenience replace emotional or social logic
+
+## Long-range drift
+
+Problem:
+
+- facts, tone, world rules, stakes, or theme slowly contradict earlier canon
+- the chapter solves the local beat but weakens the book's larger pressure system
+- marathon drafting keeps moving without re-grounding in truth files
+
+What to protect:
+
+- re-anchor every chapter in canonical files, not session momentum
+- check that new scenes still fit the novel's promise, pressure ladder, and emotional direction
+- repair contradiction immediately instead of carrying it forward
+
+## Repetition and formulaic language
+
+Problem:
+
+- sentences, scene turns, emotional beats, or chapter shapes start echoing themselves
+- stock phrasing, recycled metaphors, or benchmark cosplay replace lived texture
+- the prose keeps functioning but stops feeling alive
+
+What to protect:
+
+- vary rhythm, evidence, and scene mechanics without losing clarity
+- reject filler that merely restates what the reader already knows
+- keep project-specific voice stronger than generic model habits
@@ -1,43 +1,62 @@
-# Fantasy Module
+# 奇幻模块

-Fantasy is strongest when the world feels inhabited rather than merely explained. Strange systems need rules, cost, and everyday texture.
+奇幻最强的时候，不是设定被解释得很多，而是世界被活出来了。异质规则、能力体系、历史残留和日常生活必须互相咬合，读者才会信这个世界真的能运转。

-## Use When
+## 适用时机

- invented rules or powers matter in the chapter
- social texture should show the world is lived in
- world information must stay coherent without dumping lore
+- 本章涉及虚构规则、魔法、血统、神器、制度或异世界常识
+- 世界观需要通过社会纹理而非说明文显形
+- 设定信息必须保持一致，又不能变成 lore dump
+- 能力、资源和权力的代价要在场景里可见

-## Core Effects
+## 覆盖子型

- coherent rules
- visible costs
- social and institutional texture
- sense of history
- everyday lived reality
- controlled information release
+- 硬规则奇幻
+- 软魔法、神话式、寓言式奇幻
+- 高奇幻、低奇幻、宫廷与政治奇幻
+- 冒险、黑暗、田园、都市与混合型奇幻

-## Chapter Use
+所有子型都适用本模块。通用不变项不是“必须公式化解释一切”，而是要有可感知的世界逻辑、代价逻辑和社会后果逻辑。

- define which world rule matters now
- show it through use, friction, or consequence
- add at least one lived-in detail when texture matters
+## 核心原则

-## Read Next
+- 规则要服务冲突，而不是取代冲突
+- 能力越强，限制、代价、门槛和后果越要清楚
+- 世界可信，不靠百科全书，靠日常细节、制度摩擦和人的选择
+- 最好的设定信息总是随着人物需求一起进入，而不是停下来讲解
+- 只展示本章需要承重的规则，其余设定可以留在背景
+- 若采用软魔法或神话式处理，也要让读者感到其叙事边界、象征方向和风险，不可完全随心所欲
+
+## 章节级操作
+
+1. 先定义本章真正相关的一条世界规则。
+2. 再决定它通过使用、摩擦、失败还是代价被看见。
+3. 在需要世界厚度时，最好给一个生活化细节，让世界显得有人在其中长期生活。
+4. 若出现新规则，立刻想清楚它会改写谁的行为和社会关系。
+5. 若规则看似例外，尽量解释例外的来源和价格；若暂时不能全解释，至少标明它仍待确认的边界。
+
+## 与其他风格混用
+
+- 和 `suspense` 混用时，世界规则要让危险更具体，不要把设定当万能解答器
+- 和 `romance` 混用时，制度、身份或能力成本可以直接变成关系障碍
+- 和 `literary` 混用时，意象、历史和设定要共同承担主题，而不是彼此分家
+
+## 继续阅读

 - `world-rules-and-cost.md`
 - `everyday-world-thickness.md`

-## Revision Checks
+## 修订检查

- are relevant rules consistent
- is cost visible where power appears
- does the world feel inhabited beyond the main cast
- does the strange element change behavior or society
+- 本章相关规则是否前后一致
+- 力量出现时，代价和限制是否可见
+- 主角之外的人是否也像真实世界居民一样生活
+- 异常规则是否已经改变了行为、礼俗、机构或市场
+- 有无整段设定说明可以改成场景化呈现

-## Common Failures
+## 常见失误

- lore dumping
- power without cost
- empty world outside the main cast
- changing rules for convenience
+- 设定倾倒
+- 力量没有代价
+- 世界只围着主角转
+- 为了方便推进情节随手改规则
@@ -1,12 +1,30 @@
-# Everyday World Thickness
+# 日常世界厚度

-Secondary worlds feel credible when daily life reflects their strange rules.
+次生世界之所以可信，不在于历史写得多，而在于普通人的日常已经被那些奇异规则塑形。

-Look for:
+## 优先观察什么

- food, labor, trade, transport
- etiquette shaped by power
- common fears and assumptions
- ordinary uses of the extraordinary
+- 食物、劳动、交易、运输如何受规则影响
+- 礼仪、阶层、称呼和禁忌如何受权力分配影响
+- 人们默认害怕什么、相信什么、误会什么
+- 超常之物在普通生活中的低配用途是什么
+- 儿童、老人、穷人、官员、商贩各自如何适应这个世界

-One grounded detail is often stronger than a paragraph of history.
+## 写作动作
+
+- 每逢需要世界厚度时，优先补一个可触摸、可执行、可交易的细节
+- 如果出现高概念设定，顺手写出它对便宜货、底层职业或地方习俗的影响
+- 让背景人物也携带世界信息，不要只有主角会谈设定
+- 把历史感藏进旧建筑、法律残片、口头习惯和物价差异里
+
+## 判断标准
+
+- 不解释宏大设定，读者也能从细节里感到这世界的运行方式
+- 细节不是景观卡片，而是和人物任务、选择或危险有关
+- 一条生活细节能反推出一条更大的制度规则
+
+## 常见误区
+
+- 只会写地图和传说，不会写普通人怎么活
+- 细节漂亮但不能承重
+- 所有背景人物都像穿着异世界皮肤的现代人
@@ -1,11 +1,32 @@
-# World Rules And Cost
+# 世界规则与代价

-Every meaningful system should answer:
+任何承担冲突解法或世界秩序的体系，不管是魔法、血统、技艺、禁忌还是神器，最好能回答这五个问题中的大部分。若是偏软、偏神话或偏寓言的体系，至少也要让读者把握其边界、代价来源或介入逻辑。

- what can it do
- what can it not do
- what does it cost
- who controls access
- what social behavior has grown around it
+## 必答五问

-If a rule changes, update the main project files after the chapter.
+- 它能做什么
+- 它不能做什么
+- 它要付出什么代价
+- 谁掌握使用权和解释权
+- 围绕它形成了什么社会行为
+
+## 写作动作
+
+- 若该体系在当前阶段承担主要冲突，尽量让读者较早看到一次真实使用和一次真实受限
+- 代价优先落在身体、资源、时间、名誉、伦理或关系上
+- 如果某能力很稀缺，就写出围绕它形成的制度、市场、歧视或崇拜
+- 若出现例外情况，尽量补上为什么是例外，以及例外是否可复制；若此刻不能说满，也要标记仍待澄清的边界
+- 本章动到规则后，要同步回写主项目文件
+- 如果体系偏软，不必把机制写成手册，但仍要交代读者该如何判断“这次介入是否合理”
+
+## 判断标准
+
+- 规则能制造麻烦，而不只是解决麻烦
+- 角色做选择时会主动考虑代价
+- 读者能从人的行为看出规则存在，而不只从解释看出
+
+## 常见误区
+
+- 只有能力，没有限制
+- 只有设定条款，没有制度后果
+- 关键时刻临时冒出新规则救场
@@ -1,42 +1,62 @@
-# Horror Module
+# 恐怖模块

-Horror works by making the familiar feel wrong and progressively harder to survive or explain. Fear should move through perception, implication, and consequence.
+恐怖不是单次惊吓，而是让熟悉之物变得不再可靠，并且这种不可靠越来越难解释、难承受、难活过。恐惧应经过感知、推断和后果，而不是只经过形容词。

-## Use When
+## 适用时机

- dread should intensify across the chapter
- ordinary reality needs destabilization
- perspective limits should sharpen disturbance
+- 不祥感需要跨章节或跨场景累积
+- 日常现实需要被逐步扭斜，或现成威胁需要被不断具体化
+- 受限视角、局部多视角或框架叙事能强化不确定和压迫
+- 惊扰之后还需要留下余波

-## Core Effects
+## 覆盖子型

- dread
- distorted normality
- sensory disturbance
- limited perspective
- escalating violation
- lingering aftereffect
+- 超自然恐怖
+- 心理恐怖
+- 哥特、民俗、宗教、宇宙恐怖
+- 身体恐怖、密闭求生、怪物型恐怖
+- 与推理、悬疑、文学混合的恐怖

-## Chapter Use
+这些子型的外壳不同，但通用支架一致：正常秩序被污染，解释逐步失效，人物承受持续后果。受限视角是强工具，不是唯一合法写法。

- decide what ordinary thing becomes wrong
- decide what the point of view can and cannot know
- make fear concrete through sensory evidence and consequence
+## 核心原则

-## Read Next
+- 先让读者感觉不对，再决定何时解释
+- 恐惧常常从普通事物的轻微错位开始，但明牌威胁、身体失控或空间困局也可以作为合法入口
+- 感官证据越具体，恐惧越成立
+- 视角越近，恐惧越私人；视角越稳，失真越有力
+- 惊吓如果不改变人物、环境或规则，只会快速蒸发
+- 场景、地点和氛围本身也能构成威胁，不要把恐怖全压给怪物出场
+
+## 章节级操作
+
+1. 先决定本章恐惧从哪里进入：日常错位、明牌威胁、身体失控、空间困局、知识污染，还是怪物逼近。
+2. 再决定角色此刻能知道什么、误解什么、来不及知道什么。
+3. 用与该入口匹配的证据把恐惧做实，例如感官异常、肉体变化、场所压迫、行为异样、时间错位或追逐痕迹。
+4. 升级时优先增加解释失效、生存代价或污染范围，而不是单纯加大声响。
+5. 收尾后留下一个持续存在的后果，让恐惧进入下一章。
+
+## 与其他风格混用
+
+- 和 `suspense` 混用时，悬念负责逼近，恐怖负责扭曲与余震
+- 和 `mystery` 混用时，异常现象也要留下可追踪证据，不能只凭氛围
+- 和 `literary` 混用时，意象可以加深压迫，但不要把可感的恐惧写成抽象寓言
+
+## 继续阅读

 - `fear-escalation.md`
 - `limited-pov.md`

-## Revision Checks
+## 修订检查

- does the scene disturb before it explains
- is fear sensory rather than abstract
- does the limited perspective deepen dread
- does the chapter leave residue after the scare
+- 场景是否先扰动读者，再解释原因
+- 恐惧是否通过感官证据成立
+- 受限视角是否让不安更锋利，而不是让读者更糊涂
+- 惊吓之后是否留下余波、损伤或规则变化

-## Common Failures
+## 常见失误

- overexplaining too early
- describing fear without evidence
- noise instead of dread
+- 解释来得太早
+- 只说“很害怕”，不给证据
+- 只有噪声，没有递进的不祥感
+- 让怪物或答案完全代替氛围和场景
@@ -1,11 +1,30 @@
-# Fear Escalation
+# 恐惧升级

-Escalate fear in stages:
+恐怖升级要像螺丝收紧，不像音量放大。最稳的路径是让角色一步步失去安全解释，再失去安全选择。

-1. slight wrongness
-2. pattern recognition
-3. loss of safe explanation
-4. concrete threat
-5. consequence that lasts after the scene
+## 常用升级阶梯

-Escalation should increase cost, not only volume.
+1. 轻微不对劲
+2. 发现这不是偶然，而是模式
+3. 原本可用的安全解释失效
+4. 威胁变得具体，角色必须应对
+5. 场景结束后仍有残留后果
+
+## 写作动作
+
+- 每次升级都让角色少掉一个安慰自己的理由
+- 让异常从感官层面进入行动层面，再进入结果层面
+- 如果要慢烧，就让“证明它真的存在”的门槛越来越低
+- 升级之后要有余波，例如伤口、污染、关系破裂、规则确认
+
+## 判断标准
+
+- 后一层比前一层不仅更响，而且更贵
+- 角色的应对方式在升级中不断失效
+- 读者会感觉到世界本身变得更不可信
+
+## 常见误区
+
+- 每一层都只是更大声、更血腥
+- 角色一直重复同一种害怕反应
+- 有冲击，没有残留
@@ -1,11 +1,30 @@
-# Limited Point Of View
+# 有限视角

-Limited perspective strengthens horror because uncertainty becomes structural.
+有限视角之所以适合恐怖，不是因为它能胡乱藏信息，而是因为它把“不知道”变成了结构本身。读者被迫和角色一起受困。

-Use it to control:
+## 重点控制什么

- what the character hears but cannot place
- what they see but cannot interpret
- what they realize too late
+- 角色听见了什么，却无法定位
+- 角色看见了什么，却来不及解释
+- 角色何时意识到自己一直理解错了
+- 角色忽略了哪些本该注意的异常

-Do not use limitation to hide basic scene logic. Use it to sharpen dread.
+## 写作动作
+
+- 尽量贴近身体感受、视线盲区、注意力偏移和错误判断
+- 保持基础场景逻辑清楚，让“不知道”集中在威胁本身，而不是空间动线
+- 如果使用第一人称或紧贴第三人称，就让主观解释成为恐惧的一部分
+- 角色一旦确认了某个事实，就不要无故继续装不知道
+- 如果使用多视角或框架叙事，也要保证每个局部视角内部仍然保有不确定和压迫，而不是一换镜头就泄光
+
+## 判断标准
+
+- 读者能准确知道角色此刻看到了什么、没看到什么
+- 视角限制让恐惧更私人，而不是让文本更混乱
+- 关键信息的缺失有明确来源：距离、黑暗、噪音、偏见、创伤、时间差
+
+## 常见误区
+
+- 用视角限制掩盖场景调度不清
+- 一味藏信息，导致读者无法投入
+- 角色已经看到事实，文本却继续硬拖
@@ -1,18 +1,36 @@
-# Callback And Escalation
+# 回调与升级

-Callbacks work when a prior beat returns with changed pressure.
+回调不是重复一遍笑点，而是让旧材料在新压力下返回。升级也不是把场面写得更吵，而是让代价越来越难收拾。

-Strong callback pattern:
+## 强回调模式

-1. plant a small comic fact
-2. repeat it in a harder context
-3. transform it into a payoff that reveals character or relationship truth
+1. 先埋一个小而清晰的滑点。
+2. 在更高压力场合让它再次出现。
+3. 让第三次出现不只是好笑，还要揭示人物、关系或局势真相。

-Escalation pattern:
+## 常见升级阶梯

-1. normal awkwardness
-2. public complication
-3. status damage
-4. emotional or plot consequence
+1. 私下尴尬
+2. 被旁人看见
+3. 地位受损或计划受阻
+4. 情感后果或情节后果落下

-Do not escalate only in noise. Escalate in cost.
+## 写作动作
+
+- 每次回返都改变至少一项条件：场合、观众、 stakes、情绪或关系位置
+- 回调前先让读者记得，不必大声提醒，只要在场景里自然带过
+- 升级时优先增加暴露风险和补救成本
+- 如果笑点已完成使命，就收，不要贪第四次第五次
+
+## 判断标准
+
+- 第二次出现比第一次更难堪或更危险
+- 第三次出现带来的是结果，不只是重复识别
+- 即使读者忘记术语“回调”，也会感觉这件事终于撞回来了
+
+## 常见误区
+
+- 只重复台词，不升级压力
+- 用旁人提醒代替自然回返
+- 一味加大音量、动作、人数，却没有增加成本
+- 明明已经落地，却舍不得收手
@@ -1,44 +1,65 @@
-# Humor Module
+# 幽默模块

-Use humor as story force, not floating joke delivery. The strongest comic chapters make pressure visible through embarrassment, bias, status damage, or callback payoff.
+把幽默当成叙事压力的转向器，不要当成悬空段子投递。真正有效的小说幽默，通常来自角色越想维持体面、权威、冷静或正确，现实越不给他面子。

-## Use When
+## 适用时机

- a chapter needs release without losing stakes
- character friction should become funny before it becomes costly
- dialogue needs sharper timing
+- 章节需要释放张力，但不能削弱 stakes
+- 人物摩擦先要显得好笑，后面才更显得有代价
+- 关系中的高低位、误判、逞强、遮掩需要被看见
+- 对话节奏需要更利落、更有回弹

-## Core Effects
+## 覆盖子型

- tension release through character behavior
- status reversal
- setup and payoff
- escalation
- callback
- compressed dialogue rhythm
+- 角色窘迫型幽默
+- 冷幽默与死正经反差
+- 关系互怼与轻喜剧节奏
+- 黑色幽默与苦中作笑
+- 观察式、场景式、设定式幽默

-## Chapter Use
+以上子型都能调用本模块。通用不变项只有一条：笑意必须改变压力、关系、地位或读者对人物的理解。

- define who becomes funniest under pressure and why
- anchor humor to desire, fear, shame, or imbalance
- keep setups short and payoffs clearer than the setup
- let callbacks reveal relationship or character truth
+## 核心原则

-## Read Next
+- 幽默最好从欲望、羞耻、偏见、嘴硬和误判里长出来
+- 最稳的笑点不是“有趣的一句”，而是“角色认真时发生了错位”
+- 笑点要么改变地位，要么暴露关系，要么制造后续成本
+- 铺垫通常应更短，回报应更清楚，余波应比包袱更重要
+- 同一个人不能在所有场景都同样好笑，幽默要有人物专属性
+- 幽默也可以来自世界观察、场景安排和叙述角度，不必全部落在出丑
+
+## 章节级操作
+
+1. 先定义本章笑意主要从哪里来：窘迫、错位、观察、关系摩擦、设定反差，还是黑色余味。
+2. 如果是人物或关系型幽默，再定义谁最怕丢脸、失控或被看穿，谁又最容易误判局势。
+3. 如果是观察式、设定式或冷幽默，先明确哪一个日常逻辑、叙述角度或世界规则会产生反差。
+4. 让笑意在最合适的时点改变场景，而不是让作者硬塞笑话。
+5. 若使用回调，必须在更难堪、更公开、更失衡或更具后果的场合回返。
+6. 如果本章还有严肃任务，笑点之后要留下关系、决定、理解偏差或代价的变化。
+
+## 与其他风格混用
+
+- 和 `romance` 混用时，优先让幽默暴露心动、逞强、误会和照顾，而不是破坏张力
+- 和 `suspense` 混用时，幽默可来自错误自信或时机错位，但不能把危险感抹平
+- 和 `literary` 混用时，幽默可以服务人物尊严与社会纹理，不要写成作者的机灵表演
+
+## 继续阅读

 - `humor-core.md`
 - `dialogue-comedy.md`
 - `callback-and-escalation.md`

-## Revision Checks
+## 修订检查

- does humor grow from character rather than hover above the scene
- is the setup shorter than the burden it creates
- do callbacks evolve instead of merely repeating
- does humor serve the chapter mission
+- 笑点是否来自人物处境，而不是作者跳出来抖机灵
+- 去掉最搞笑的那一句后，场景是否仍然推进了关系或情节
+- 铺垫是否足够短，回报是否足够清晰
+- 回调是否带来了升级，而不是原样再说一遍
+- 主要人物的笑法是否彼此不同

-## Common Failures
+## 常见失误

- writing stand-up instead of fiction
- turning everyone into the same joke machine
- sacrificing emotional continuity for a cheap gag
+- 把小说写成脱口秀
+- 每个人都像同一个嘴贫角色
+- 为了笑点破坏人物情绪连续性
+- 只把场面写得更吵，却没有让代价更高
@@ -1,18 +1,34 @@
-# Dialogue Comedy
+# 对话幽默

-Comedy dialogue should feel faster and cleaner than ordinary dialogue, but still belong to the speaker.
+好笑的对话通常比普通对话更短、更快、更干净，但它仍然必须属于说话的人，而不是属于作者。

-Use:
+## 常用手法

- short setups
- delayed admission
- selective misunderstanding
- over-literal response
- under-response to a dramatic moment
+- 短铺垫，长后劲
+- 该承认的事延迟承认
+- 有选择地误解对方
+- 对情绪性发言给出过度字面化的回应
+- 面对重大时刻故意低配反应
+- 把真正想说的东西藏在绕路表达里

-Check:
+## 写作动作

- does each speaker sound like themselves
- can the line be shortened
- is the funny beat placed on the strongest last phrase
- is the subtext doing part of the work
+- 尽量让笑点落在句末最强的位置
+- 一次对答里只保留一个最有效的滑点，不要每句都抖包袱
+- 如果人物本来寡言，就让他的幽默靠删减和错位，不靠话密
+- 如果是双人对话，想清楚谁在主动控制节奏，谁在被动接招
+- 用停顿、打断、否认和改口制造节拍，而不是堆形容词
+
+## 判断标准
+
+- 每个说话者都仍然像自己
+- 这句台词还能再短一点
+- 好笑的部分同时暴露了态度、关系或防御机制
+- 即使读者没大笑，也能感觉到人物之间的电流和节奏
+
+## 常见误区
+
+- 所有人说话都变成一个机灵模板
+- 明明情绪正在升高，却用过多段子打断
+- 笑点藏在句中，句尾反而虚
+- 用解释代替节奏
@@ -1,24 +1,39 @@
-# Humor Core
+# 幽默核心

-In fiction, humor works best when it emerges from pressure.
+小说里的幽默，最稳的引擎不是聪明，而是压力。角色越认真地想守住某件事，幽默就越容易成立。

-Useful engines:
+## 常用引擎

- a character trying too hard to seem controlled
- a mismatch between self-image and reality
- status inversion
- repeated avoidance that becomes impossible
- overconfident interpretation meeting hard fact
+- 角色拼命装稳，现实却持续拆台
+- 自我认知与外部事实不匹配
+- 地位翻转，强者暂时失手，弱者意外占上风
+- 反复回避的问题终于无法回避
+- 过度自信的解释撞上冷硬事实
+- 一件小小失误被关系、场合或围观者放大

-Prefer:
+## 写作动作

- character-specific humor
- relational humor
- scene-based embarrassment
- compressed observation
+- 给笑点绑定一个明确的人物弱点，例如面子、洁癖、傲慢、好胜、占有欲
+- 先让角色认真行动，再让偏差显形；不要先想 punchline 再补场景
+- 保留人物自己的说话方式，别为了搞笑把人写串味
+- 让笑点和下一步行动相连，最好能逼出补救、掩饰或更大的失误

-Avoid:
+## 判断标准

- unrelated one-liners
- funny prose that erases scene logic
- humor that destroys the novel's emotional continuity
+- 读者会觉得意外，但回看又觉得必然
+- 这个笑点换到另一个人物身上会变弱，因为它依赖此人的性格缺口
+- 笑完之后，关系、地位或成本比之前更明确
+
+## 优先选择
+
+- 人物专属的幽默
+- 关系里的幽默
+- 场景式窘迫
+- 压缩而具体的观察
+
+## 避免
+
+- 与场景任务无关的一次性俏皮话
+- 用“有趣的叙述腔”淹没场景逻辑
+- 靠羞辱弱者取笑却没有人物和情境支撑
+- 用幽默抹掉本来该疼的情绪后果
@@ -1,67 +1,76 @@
-# Internal Style Modules Index
+# 内部风格模块索引

-Use this file to route style loading inside `novel-control-station`.
+用这个索引在 `novel-control-station` 内部做风格路由，不要凭印象混读全部模块。

-Rule:
+## 总规则

- load only the selected modules
- read each module's `core.md` first
- read deeper files only when the chapter needs them
- keep at most 2 primary styles and 1 support style active
+- 只加载当前章节真正激活的风格模块
+- 永远先读对应模块的 `core.md`
+- 只有当章节压力已经明确时，才继续读取更深文件
+- 同时保持最多 2 个主风格和 1 个辅助风格
+- 风格模块服务于章节任务，不要把它们当成装饰性气味包

-## Humor
+## 选择顺序

- use when the chapter needs comedy from pressure, embarrassment, status shift, or callback timing
- read first: `humor/core.md`
- drill deeper when needed:
-  - `humor/humor-core.md`
-  - `humor/dialogue-comedy.md`
-  - `humor/callback-and-escalation.md`
+1. 先判断章节主要靠什么推进。
+2. 再判断读者此章最该感到什么。
+3. 最后判断哪一个辅助风格能增强，不是稀释，当前主线压力。

-## Suspense
+如果一个模块只是让文本“更像某种类型”而不能改变压力、关系、风险或意义，就不要加载它。

- use when the chapter needs uncertainty, danger pressure, delayed revelation, or a forward-driving hook
- read first: `suspense/core.md`
- drill deeper when needed:
-  - `suspense/suspense-core.md`
-  - `suspense/chapter-hook.md`
+## Humor / 幽默

-## Mystery
+- 适用于章节需要由窘迫、失衡、观察反差、冷幽默、回调或说话节奏制造笑意
+- 先读：`humor/core.md`
+- `humor/humor-core.md`
+- `humor/dialogue-comedy.md`
+- `humor/callback-and-escalation.md`

- use when the project needs investigable questions, fair clues, suspect control, or staged revelation
- read first: `mystery/core.md`
- drill deeper when needed:
-  - `mystery/clue-and-misdirection.md`
-  - `mystery/reveal-order.md`
+## Suspense / 悬念

-## Romance
+- 适用于章节需要维持未知、危险、暴露风险、延迟揭示或强前拉钩子
+- 先读：`suspense/core.md`
+- `suspense/suspense-core.md`
+- `suspense/chapter-hook.md`

- use when attraction, emotional debt, unresolved intimacy, or relationship temperature change drives the chapter
- read first: `romance/core.md`
- drill deeper when needed:
-  - `romance/emotional-debt.md`
-  - `romance/subtext-and-tension.md`
+## Mystery / 推理

-## Horror
+- 适用于项目需要可调查问题、公平线索、可能性分支、误导和分段揭晓
+- 先读：`mystery/core.md`
+- `mystery/clue-and-misdirection.md`
+- `mystery/reveal-order.md`

- use when dread, destabilized normality, sensory fear, or limited-perspective disturbance drives the chapter
- read first: `horror/core.md`
- drill deeper when needed:
-  - `horror/fear-escalation.md`
-  - `horror/limited-pov.md`
+## Romance / 爱情

-## Fantasy
+- 适用于章节核心由吸引、阻碍、情感债、未决亲密或关系温度变化驱动
+- 先读：`romance/core.md`
+- `romance/emotional-debt.md`
+- `romance/subtext-and-tension.md`

- use when invented systems, world rules, visible cost, or lived-in secondary-world texture matter
- read first: `fantasy/core.md`
- drill deeper when needed:
-  - `fantasy/world-rules-and-cost.md`
-  - `fantasy/everyday-world-thickness.md`
+## Horror / 恐怖

-## Literary
+- 适用于章节核心由不祥感、明牌威胁、身体或空间压迫、感官惊扰、视角限制和余波构成
+- 先读：`horror/core.md`
+- `horror/fear-escalation.md`
+- `horror/limited-pov.md`

- use when thematic pressure, motif return, language discipline, or formal intensity needs reinforcement
- read first: `literary/core.md`
- drill deeper when needed:
-  - `literary/theme-pressure.md`
-  - `literary/language-health.md`
+## Fantasy / 奇幻
+
+- 适用于章节需要处理虚构规则、神话性介入、权力代价、制度纹理、异世界日常和设定一致性
+- 先读：`fantasy/core.md`
+- `fantasy/world-rules-and-cost.md`
+- `fantasy/everyday-world-thickness.md`
+
+## Literary / 文学
+
+- 适用于章节需要强化主题压力、意象回返、语言纪律、结构意味和解释层次
+- 先读：`literary/core.md`
+- `literary/theme-pressure.md`
+- `literary/language-health.md`
+
+## 混用提醒
+
+- `suspense + mystery`：前者管悬而未决的压力，后者管证据与答案的公平性
+- `romance + humor`：优先让笑点暴露亲密风险或关系差值，不要把情绪债轻易消解掉
+- `horror + suspense`：恐怖更重失真和余波，悬念更重未知与逼近，二者通常能叠加，但不要混成一团
+- `fantasy + literary`：最好让设定与语言都承担意义，避免一个只负责说明、一个只负责装饰
@@ -1,43 +1,62 @@
-# Literary Module
+# 文学模块

-Literary intensity comes from pressure between language, character, and theme. Depth should arise through scene, image, and choice rather than direct sermon.
+文学性不是高腔调，而是语言、人物和主题在场景里互相施压。深度必须通过选择、后果、意象回返和结构张力生长出来，而不是靠作者直接宣布意义。

-## Use When
+## 适用时机

- thematic pressure needs reinforcement
- image or motif return should carry meaning
- language discipline or formal intensity matters
+- 章节需要强化主题压力
+- 某个意象、母题、场景回声或形式安排需要承担意义
+- 语言纪律、叙述密度、心理深描或结构性对照需要提升
+- 项目明确要求偏文学、克制、实验或高解释层次

-## Core Effects
+## 覆盖子型

- thematic pressure
- image and motif return
- precise language
- psychological depth
- meaningful structural movement
- layered interpretation
+- 现实主义文学
+- 抒情、冷峻、极简、繁密、碎片化写法
+- 心理深描、社会观察、形式实验
+- 与悬疑、奇幻、爱情等类型混合的文学化写法

-## Chapter Use
+本模块不是在规定单一文风。通用不变项只有这些：形式要承担意义，语言要承重，人物与主题要互相施压。

- define what thematic pressure the chapter applies
- know which motif or image should recur, if any
- use structural complexity only when it reveals character truth or thematic relation
+## 核心原则

-## Read Next
+- 主题应压在人物身上，不应飘在人物上方
+- 意象和母题要随着情境变化而变义，不要只做漂亮回环
+- 语言的精确度比语言的华丽度重要
+- 结构复杂只有在揭示人物真相、时间关系或意义冲突时才值得使用
+- 含混可以有，但必须是多义，不是糊
+- 文风可以繁、可以简，但都必须让读者更接近经验，而不是更接近作者姿态
+
+## 章节级操作
+
+1. 先写清本章到底在逼问什么价值冲突。
+2. 再决定哪一个动作、意象或结构回返来承重。
+3. 把抽象命题拆回人物选择和后果。
+4. 修辞出现时问一句：它除了好看，还做了什么场景工作。
+5. 如使用跳时、重叠视角或非线性，应让它揭示新的心理或意义层；若只是调度便利，通常不值得上这种结构。
+
+## 与其他风格混用
+
+- 和 `romance` 混用时，主题不能替代互动，必须落到关系选择
+- 和 `mystery` 混用时，象征和含义不能破坏线索清晰度
+- 和 `fantasy` 混用时，世界观细节与主题意象应相互支撑
+
+## 继续阅读

 - `theme-pressure.md`
 - `language-health.md`

-## Revision Checks
+## 修订检查

- does the chapter press a real thematic conflict
- do images carry weight rather than decoration
- is the language alive instead of inflated
- does structure reveal character truth
+- 章节是否真的压出了一个价值冲突
+- 意象是否承重，而非装饰
+- 语言是否活着，而不是膨胀着
+- 结构是否揭示人物真相或主题关系
+- 评论是否已经取代了戏

-## Common Failures
+## 常见失误

- fake depth through abstraction
- dead language
- nonlinear movement without purpose
- commentary replacing drama
+- 用抽象句制造假深度
+- 语言死、空、肿
+- 非线性只是花样
+- 作者评论挤掉人物经验
@@ -1,15 +1,41 @@
-# Language Health
+# 语言健康

-Watch for dead language:
+语言一旦死掉，文学性就会立刻露馅。最常见的死法不是语病，而是抽象膨胀、假深刻、空修辞和没有场景工作的漂亮句子。

- inflated abstraction
- slogan-like summary
- fake-profound declarations
- decorative sentences with no scene work
+## 重点排查

-Prefer:
+- 抽象名词堆积
+- 口号式总结
+- 伪深刻判断句
+- 只好看、不承重的修辞
+- 所有句子长度和力度都一个样

- specific nouns and verbs
- image with consequence
- syntax that matches emotional temperature
- clarity before ornament
+## 优先保留
+
+- 具体名词和有方向的动词
+- 会带来后果的意象
+- 和情绪温度匹配的句法节奏
+- 先清楚，再精致
+- 必要时保留有控制的繁复、跳接和抒情，但前提是它们确实在制造经验和意义
+- 有意为之的空白、摘要、议论或碎片段，只要它们确实承重，也不必硬改成感官描写
+
+## 写作动作
+
+- 把“他感到一种难以言说的……”之类句子压回到身体、动作、环境和记忆触发
+- 在承重段落里，尽量留一个真正能被看到、听到、摸到或误认的细节；摘要、议论或碎片段不必机械补齐
+- 在强情绪段落中主动调整句长，避免整页同拍
+- 遇到漂亮句时问一句：删掉它，场景损失了什么；如果答案是“只损失好看”，就重写
+
+## 判断标准
+
+- 语言能让读者更接近人物经验，而不是更远
+- 描写不是悬空标签，而是能改变理解和情绪
+- 哪怕文风克制，文本也依然有生命脉搏
+
+## 常见误区
+
+- 把模糊当高级
+- 把浓度当深度
+- 把漂亮话当人物感受
+- 全文都像同一种速度、同一种亮度
+- 为了“去 AI 味”过度削平文本，反而把应有的文体力度也删没
@@ -1,11 +1,28 @@
-# Theme Pressure
+# 主题压力

-Theme should not float above the scene. It should pressure the scene.
+主题不是章节的口号，而是章节里哪一种价值正在被逼着接受检验。只要还没有逼到选择和后果，主题就还停留在口头层。

-Useful questions:
+## 三个关键问题

- what value is being tested here
- what contradiction is being exposed
- what choice forces the chapter's meaning into action
+- 这一章到底在测试什么价值
+- 哪个矛盾正在被暴露
+- 哪个选择把意义逼进了行动

-If the chapter has a stated theme but no pressured choice, the theme is probably ornamental.
+## 写作动作
+
+- 让不同人物各自代表并扭曲一部分主题，而不是让一个人负责说理
+- 把主题问题压进具体局面，例如要不要背叛、要不要沉默、要不要承担
+- 让主题落在后果上，人物做完选择后必须有代价或残留
+- 如果使用母题或意象，让它在不同章节里承担不同情绪和道德重量
+
+## 判断标准
+
+- 不提主题词，读者也能感觉到这一章在争什么
+- 主题冲突改变了人物关系或自我认知
+- 主题不是把场景解释清楚，而是让场景更有压力
+
+## 常见误区
+
+- 角色突然说出一大段正确观点
+- 主题与人物选择脱节
+- 只有立场，没有代价
@@ -1,15 +1,31 @@
-# Clue And Misdirection
+# 线索与误导

-Fair-play mystery does not require obvious clues. It requires real clues.
+公平推理不要求线索显眼，但要求线索真实存在。读者不一定立刻认出它是关键，但回看时必须承认它一直在那里。

-Good clue practice:
+## 好线索的基本做法

- place clues in ordinary scene business
- let characters misinterpret them for believable reasons
- let a red herring explain the clue too neatly
+- 把线索放进正常场景运作里，而不是单独摆在聚光灯下
+- 让人物因为偏见、时间压力、知识不足或情感立场而误读它
+- 让红鲱鱼给出一个过于顺手、但暂时足够成立的解释
+- 让同一条线索同时服务两个层面：表层解释和真相解释

-Bad clue practice:
+## 写作动作

- hiding the clue entirely
- revealing the clue only in exposition later
- making every suspect obviously fake
+- 每条真线索至少回答三个问题：谁能看到、谁会误解、何时会回响
+- 如果线索很细，补一个能让读者记住它的场景动作或物理特征
+- 红鲱鱼最好也推动人物采取行动，而不是只当烟幕
+- 让主要错误解释各自带出不同误导路径；如果是嫌疑人池结构，再避免所有嫌疑人都像同一种设计
+- 不是每一条假线索都必须在结尾逐项解释，只有真正承重、足以让读者索求答案的那类误导才需要回收
+
+## 判断标准
+
+- 线索在第一次出现时足够自然，但第二次回看会变亮
+- 误读的理由来自人物和情境，而不是作者单纯不让人看见
+- 红鲱鱼在被推翻前看起来真的可行
+
+## 坏线索的典型形态
+
+- 关键线索根本没出现
+- 关键线索只在后期解释段里被补写
+- 所有嫌疑人都明显是陪跑
+- 为了误导而让人物做不合逻辑的蠢事
@@ -1,41 +1,62 @@
-# Mystery Module
+# 推理模块

-Mystery depends on controlled evidence. The reader may be misled, but the truth must be discoverable in retrospect.
+推理的根不是故弄玄虚，而是证据管理。读者可以被带偏，但真相必须在回看时可追溯、可复盘、可自证。

-## Use When
+## 适用时机

- a central question should be genuinely investigable
- clues and suspects need tight control
- revelation order matters as much as the answer
+- 中心问题需要被真正调查，而不是等作者宣布答案
+- 线索、可能性分支、错误解释需要同步控制
+- 真相的揭晓顺序和真相本身同样重要
+- 读者需要在“能猜”和“猜不透”之间维持张力

-## Core Effects
+## 覆盖子型

- investigable central question
- fair-play clues
- viable suspects or possibilities
- controlled misdirection
- layered reveal order
+- whodunit
+- howdunit
+- whydunit
+- 本格、公路、程序、轻推理、黑色混合型
+- 带超自然或幻想元素的调查叙事

-## Chapter Use
+各子型都适用本模块。通用硬约束是证据必须真实存在，结尾必须能回推，误导必须公平。

- know what the reader may notice
- know what is visible but likely to be misread
- ensure the final answer can be traced backward
+## 核心原则

-## Read Next
+- 中心问题必须具体，最好能落到谁、为什么、如何、何时中的一个或多个
+- 真线索应该尽早出现，但可以被合理误读
+- 红鲱鱼的作用不是作弊，而是提供一个看起来更顺手的错误解释
+- 如果该子型依赖嫌疑人池或多种可能性分支，它们都要各自成立，而不是只有一个真选项像活人
+- 结局必须能倒推回来，而不是靠最后一章空降信息
+
+## 章节级操作
+
+1. 先明确本章新增或重估的证据是什么。
+2. 再明确这份证据此刻最合理的误读方式是什么。
+3. 让人物对证据做出不同解读，推动关系、判断和行动变化。
+4. 如果本章不揭晓真相，也要推进调查位置，而不是原地兜圈。
+5. 每次揭晓后都检查是否还保留足够悬而未决的调查价值。
+
+## 与其他风格混用
+
+- 和 `suspense` 混用时，推理控制真相路径，悬念控制逼近感和时间感
+- 和 `horror` 混用时，超自然或异常现象也要遵守证据公平，不能只靠气氛顶替逻辑
+- 和 `literary` 混用时，象征和主题可以加深意味，但不能模糊证据链
+
+## 继续阅读

 - `clue-and-misdirection.md`
 - `reveal-order.md`

-## Revision Checks
+## 修订检查

- are true clues present early enough
- do red herrings divert without cheating
- can the answer be reverse-engineered
- does mystery pressure alter character behavior
+- 真线索是否出场得足够早
+- 红鲱鱼是否有其自身合理性，而不是作者硬拐
+- 读者是否能在结尾回推答案
+- 调查压力是否改变了人物行为和关系
+- 本章是否真的让“知道更多”或“误得更深”

-## Common Failures
+## 常见失误

- introducing the answer too late
- making clues invisible instead of misreadable
- unloading all answers at once
+- 到最后才引入关键答案
+- 线索不是被误读，而是被彻底藏掉
+- 所有主要错误解释都明显站不住，或所有嫌疑人都明显陪跑
+- 一次性把全部答案用说明文倒出来
@@ -1,12 +1,29 @@
-# Reveal Order
+# 揭晓顺序

-Do not unload the answer in one block if staged revelation is stronger.
+推理的揭晓往往不是“宣布答案”，而是逐层拆除旧解释。顺序安排得好，真相会像结构性坍塌一样出现，而不是像老师公布标准答案。

-Useful order:
+## 常用顺序

-1. remove the wrong explanation
-2. expose the hidden pattern
-3. connect motive, method, and timing
-4. show the cost of the truth
+1. 先推翻最顺手的错误解释。
+2. 再显出此前被忽略的模式。
+3. 把动机、手法、时机和机会连接起来。
+4. 最后展示真相带来的情感、关系或制度代价。

-The later the reveal, the more earlier clues must bear weight.
+## 写作动作
+
+- 不要等到最后一分钟才给读者理解真相所需的关键拼图
+- 每次小揭晓都应改变怀疑方向、人物站位或行动策略
+- 如果答案很晚出现，前面的线索就必须更能承重
+- 真相出来后不要立刻收尾，留出“知道之后怎么办”的压力
+
+## 判断标准
+
+- 揭晓之后，读者会想回翻前文，而不是抱怨信息不足
+- 每一层揭晓都把局面推向更难处理的位置
+- 真相不仅解释了事件，也重写了人物关系或道德评价
+
+## 常见误区
+
+- 一口气说明全部答案，缺少层次
+- 还没给足证据就让角色突然顿悟
+- 只揭晓“谁做的”，却没让“为什么”和“代价”落地
@@ -1,42 +1,61 @@
-# Romance Module
+# 爱情模块

-Romance works when desire, fear, pride, tenderness, and emotional debt all matter. Relationship progress must come through pressure, not empty delay.
+爱情线不是靠“终于在一起”才成立，而是靠欲望、恐惧、自尊、照顾、误判和情感债在场景里持续摩擦。关系推进必须经由压力，而不是靠空转拖延。

-## Use When
+## 适用时机

- attraction and obstacle shape the chapter
- relational tension needs earned movement
- intimacy should appear through behavior rather than explanation
+- 吸引和阻碍共同塑造章节
+- 关系张力需要有代价地移动
+- 亲密感应通过行动、克制或遗漏显形
+- 本章需要改变关系温度，而不是重复同一种暧昧

-## Core Effects
+## 覆盖子型

- attraction with obstacle
- emotional debt
- subtext
- push-pull tension
- earned intimacy
- relationship temperature change
+- 当代、历史、校园、职场、奇幻、科幻、悬疑混合爱情
+- 慢热、敌对转爱、二次机会、禁忌关系、黑暗爱情
+- 主线爱情与副线爱情

-## Chapter Use
+各子型都适用本模块。通用不变项是：双方必须彼此重要，障碍必须真实，关系必须移动；若爱情是主风格，还需要在 `09-style-guide.md` 中写明关系推进与情感兑现预期。

- define what each side wants and fears
- decide whether the chapter brings them closer, farther, or into riskier entanglement
- let omission, restraint, or practical action carry feeling
+## 核心原则

-## Read Next
+- 化学反应来自具体的人物差异，不来自空泛赞美
+- 最稳的爱情张力是既想靠近，又怕失去、暴露或欠债
+- 多数承重关系戏最好改变一项：信任、希望、占有欲、脆弱度、边界或误会深度；较静的场景也可以通过加深既有状态来前进
+- 延迟可以有，但最好带来新压力、债务、理解、确认或代价，避免只是拖
+- 最能打动人的亲密往往藏在小动作、照顾、判断失手和说不出口里
+- 吸引之外还要有互补、互相理解或彼此改变的理由，否则只剩表面热度
+- 边界、同意和风险感要清楚，尤其在高压、禁忌或黑暗关系里
+
+## 章节级操作
+
+1. 明确双方此章各自想要什么、怕什么。
+2. 决定本章是拉近、拉远，还是把两人拖进更危险的纠缠。
+3. 让一个实际动作承担情感含义，例如帮忙、隐瞒、袒护、退让、失控。
+4. 若设置误会，优先让它来自人物认知、处境和伤口，不要靠低级不沟通。
+5. 结尾留下新的情感债、风险或期待，而不是原地踏步。
+
+## 与其他风格混用
+
+- 和 `humor` 混用时，笑点应暴露心动、防御、吃醋或亲密失衡
+- 和 `suspense` 混用时，可把告白时机、关系暴露或失去对方变成风险轴
+- 和 `literary` 混用时，主题和意象可以加深关系，但不能替代具体互动
+
+## 继续阅读

 - `emotional-debt.md`
 - `subtext-and-tension.md`

-## Revision Checks
+## 修订检查

- is relationship pressure specific rather than generic
- does the scene change relationship state
- is emotional movement visible in action or omission
- does attraction coexist with risk
+- 关系压力是否具体到人，而不是泛泛“很有感觉”
+- 场景是否改变了关系状态
+- 情绪移动是否体现在动作、遗漏或失控里
+- 吸引是否始终伴随风险、代价或边界问题

-## Common Failures
+## 常见失误

- delaying progress without new pressure
- replacing chemistry with generic praise
- using convenience misunderstandings instead of character logic
+- 一拖再拖却没有新压力
+- 把化学反应写成外貌夸奖和标签赞美
+- 误会全靠低效沟通
+- 关系戏只会升温，不会出现反噬和后果
@@ -1,13 +1,30 @@
-# Emotional Debt
+# 情感债

-Track what one character owes another emotionally.
+情感债是爱情线的蓄压器。哪怕情节暂时平缓，只要有人欠着、忍着、装着、没还上，场景就还会带电。

-Examples:
+## 常见债务类型

- withheld truth
- sacrifice not repaid
- tenderness not acknowledged
- insult not answered
- trust granted under pressure
+- 隐瞒了应该说出的真相
+- 接受了对方的牺牲却还不起
+- 收到温柔却没有回应
+- 说了伤人的话却没真正处理
+- 在压力下得到信任，却配不上那份信任

-Debt is useful because it keeps scenes charged even when the plot is temporarily calm.
+## 写作动作
+
+- 记清是谁欠谁，欠的是什么，何时形成，何时被提醒
+- 债务不一定立刻爆发，但应持续影响说话方式、判断和边界
+- 一次看似小的照顾，只要发生在正确压力下，也可以形成很重的债
+- 债务若长期不处理，利息应体现为嫉妒、回避、过度补偿或更高期待
+
+## 判断标准
+
+- 不用提“我欠你”，读者也能感觉到不平衡
+- 情感债会改变人物下一次的选择
+- 债务被触碰时，场景电流明显上升
+
+## 常见误区
+
+- 把情感债写成单向道德账本，没有关系张力
+- 债务形成了却长期没有回响
+- 只记大牺牲，不记小而关键的体贴、隐瞒和偏袒
@@ -1,18 +1,35 @@
-# Subtext And Tension
+# 潜台词与张力

-Romance scenes often work better when the literal topic is not the real topic.
+很多有效的爱情场景里，人物嘴上在谈别的，真正交换的是试探、照顾、占有、退让和脆弱度。

-Useful moves:
+## 常用动作

- speak around the feeling
- let care appear as irritation or control
- let a practical action carry emotional meaning
- let one person ask for something smaller than what they really want
+- 绕着感觉说话，不直接命名感觉
+- 让照顾伪装成不耐烦、控制欲或规矩
+- 让一个具体动作承担情绪，例如递水、挡人、送回家、删消息、停顿
+- 可以让人物只开口要一个更小的东西，实际想得到的是更大的确认
+- 让双方暴露程度不对称，从而制造张力

-Check whether the chapter changes:
+## 写作动作

- trust
- vulnerability
- jealousy
- hope
- restraint
+- 先写清 literal topic，再写清 real topic
+- 若该场景依赖潜台词推进，可以安排一个人物差点说穿又收回的瞬间
+- 用身体反应、动作选择和注意力偏移补足未说出口的内容
+- 如果张力要升，不一定靠拥抱或亲吻，也可以靠更危险的诚实
+- 如果场景已经涉及明确边界、拒绝或同意，潜台词不能拿来抹糊这些事实
+
+## 检查本章是否改变了
+
+- 信任
+- 脆弱度
+- 嫉妒
+- 希望
+- 克制
+- 边界
+
+## 常见误区
+
+- 把潜台词写成读者根本看不懂的空白
+- 所有人一到感情戏就只会说漂亮话
+- 只有吸引，没有防御和风险
+- 只靠肢体接触制造张力，忽略语言和选择
@@ -1,20 +1,35 @@
-# Chapter Hook
+# 章节钩子

-A good hook should:
+好的章节钩子不是“停在最吓人的地方”，而是让一个已有线索在结尾突然变得更贵、更急或更难回头。

- emerge from the line already in motion
- sharpen a question, threat, or irreversible choice
- make the next chapter feel necessary
+## 钩子应当做到

-Useful hook forms:
+- 从已经在动的情节线里长出来
+- 把问题、威胁或选择再拧紧一圈
+- 让下一章有明确继续动力；若项目是 slow-burn、克制或文学向，也至少要留下低烈度余压或未完成动作

- discovery with consequence
- decision under time pressure
- unexpected reversal of trust
- revealed cost of an earlier action
+## 常见钩子形态

-Avoid:
+- 发现，并且这个发现立刻带来后果
+- 时间压力下的决定
+- 信任对象突然翻面或不再可靠
+- 之前动作的真实代价此刻显形
+- 以为安全的解释被推翻

- random jump scare
- arbitrary withheld sentence
- fake cliffhanger with no actual change
+## 写作动作
+
+- 结尾前确认局面真的发生了变化，不只是多留了一个问号
+- 优先让钩子落在行动后的余震上，而不是空白信息上
+- 如果本章是慢悬念，可以只让读者看见下一步麻烦的轮廓，或确认一股尚未解决的余压
+
+## 判断标准
+
+- 这一结尾能否一句话说出“为什么现在更糟了”
+- 如果把结尾挪到上一段，是否会损失明显力量
+- 下一章最自然的开头是否已经被迫确定
+
+## 避免
+
+- 随机跳惊
+- 故意不写完一句话
+- 没有实际变化的假 cliffhanger
@@ -1,43 +1,64 @@
-# Suspense Module
+# 悬念模块

-Use suspense to keep forward pull clear. The reader should know what danger, question, or irreversible cost is in motion even when the answer is delayed.
+悬念的核心不是“先不说”，而是让读者清楚地感到某个问题、危险或代价正在逼近，同时答案还没到手。

-## Use When
+## 适用时机

- uncertainty must stay live across scenes
- risk needs escalation before a reveal
- the chapter ending needs real forward pull
+- 未知必须跨场景持续发热
+- 揭示之前，风险需要先升级
+- 章节末尾需要真实的前拉力
+- 人物知道得不够，读者又比人物多一点或少一点

-## Core Effects
+## 覆盖子型

- active unanswered question
- time pressure
- information imbalance
- danger escalation
- delayed revelation
- hook-driven endings
+- 惊悚式高压悬念
+- 心理悬念
+- 慢烧悬念
+- 生存与逃亡型悬念
+- 阴谋、调查、家庭或关系悬念

-## Chapter Use
+不管子型如何变化，悬念的通用骨架都不变：有活问题，有拖延代价，有不断增多的麻烦。

- define the chapter's main uncertainty
- decide what cannot be revealed yet and why
- increase risk, cost, or exposure before the end
- make the hook emerge from an existing line
+## 核心原则

-## Read Next
+- 多数章节优先有一个主悬而未决问题，其余问题为它服务；多线或慢烧章节也可以保留多个低频问题，但最好让一条牵引最清楚
+- 延迟揭示必须有理由，例如时机未到、视角受限、代价更大、判断可能出错
+- 悬念靠“问题越来越贵”成立，不靠“信息越来越少”成立
+- 场景本身必须清楚，不能用读者看不懂来伪装悬念
+- 每个场景最好让旧问题更尖，或新增一个有分量的新问题
+- 对手、环境和偶发干扰通常应持续加码；在 slow-burn 或克制模式里，也可以只做低烈度推进，不必每场都猛拧
+- 如果 `09-style-guide.md` 已下调 hook density 或 cliff severity，章节牵引可以写成余压、未完成动作或不安确认，而不必强做 cliffhanger
+
+## 章节级操作
+
+1. 明确本章最重要的未知是什么。
+2. 写清如果角色现在弄错或来不及，会失去什么。
+3. 决定读者、角色、对手三方的信息差如何排列。
+4. 多数情况下在章节后半段让选择更窄、代价更清或局面更急，不要让角色长期原地猜。
+5. 章节结尾用现有情节线自然长出钩子，而不是外插惊吓。
+
+## 与其他风格混用
+
+- 和 `mystery` 混用时，悬念负责“现在怎么办”，推理负责“真相到底是什么”
+- 和 `horror` 混用时，悬念更偏逼近感，恐怖更偏失真感和余波
+- 和 `romance` 混用时，可把关系暴露、错失时机或情感代价变成悬念来源
+
+## 继续阅读

 - `suspense-core.md`
 - `chapter-hook.md`

-## Revision Checks
+## 修订检查

- is the main uncertainty clear
- does pressure rise rather than stall
- is delayed information justified by story logic
- does the hook create necessity, not noise
+- 主未知是否一句话就能说清
+- 风险是否持续上升，而不是原地踏步
+- 延迟信息是否有故事逻辑，而不是作者强压
+- 每一场戏结束时，问题是否更尖或成本是否更高
+- 结尾钩子是否让下一章变得必要

-## Common Failures
+## 常见失误

- vagueness mistaken for suspense
- hiding basic scene clarity
- repeating the same danger level
+- 把含糊当成悬念
+- 为了隐瞒而隐瞒，连基础动作都写不清
+- 连续重复同一级别的危险
+- 章节结尾只留一句没说完的话，没有真实变化
@@ -1,18 +1,34 @@
-# Suspense Core
+# 悬念核心

-Suspense needs a live question and a cost for delay.
+悬念要同时满足两件事：有活着的问题，也有拖延的代价。少一个，读者都会松掉。

-Core levers:
+## 五个核心杠杆

- what is about to happen
- who is exposed
- what the character does not know
- what the reader partially knows
- how long the character has before loss becomes irreversible
+- 接下来可能发生什么
+- 谁正在暴露，暴露到什么程度
+- 角色还不知道什么
+- 读者部分知道什么，或被引向什么判断
+- 距离不可逆损失还有多久

-Every suspense chapter should do at least one:
+## 常用推进方式

- deepen uncertainty
- narrow options
- increase cost
- expose a new danger
+- 让不确定性更具体，而不是更模糊
+- 让选择更少，而不是让信息原地打转
+- 让代价更高，例如时间更少、见证者更多、误判后果更重
+- 暴露一个新的危险或新的误读路径
+- 给角色一个暂时看似可行、其实更糟的选择
+- 给主角再加一层必须同时处理的麻烦，让局面更像多盘并转而不是单点冲刺
+- 让对手或外部世界反应，而不是只有主角单边行动
+
+## 写作动作
+
+- 进入场景前先写下“本场的悬念句”
+- 中段安排一次压力变化，让角色不得不修正判断
+- 后段避免纯解释，尽量用行动、被打断、失败尝试来制造张力
+- 如需延迟答案，先交代为什么现在拿不到，而不是硬拖
+
+## 判断标准
+
+- 读者能说出自己在等什么
+- 角色不是被动挨等，而是在带着不完整信息行动
+- 场景结束后，局面更窄、更险或更急