pre-write-checklist

Mandatory Pre-Writing Checklist

Core Function

Solving the AI Long-Form Loss of Focus Problem - This is the core innovation of Novel Writer Skills v1.0.

Root of the Problem

User feedback: When creating with novel-writer, the first 30 chapters are of good quality, but after chapter 30 the AI begins to:

Forget earlier settings
Character personalities become inconsistent
Plot repeats or contradicts itself
Ignore the principles of the creative constitution

Root Cause: Long conversations cause the AI to forget early context; no matter how detailed specification.md is written, it will be forgotten.

Solution

Mandatorily reread all key files before each writing session → AI reloads the complete context → maintains consistency

9-Item Mandatory Checklist

Each time/writecommand is executed, this checklist must first be completed:

📋 写作前检查清单（必须完成）：

✓ 1. memory/constitution.md - 创作宪法
✓ 2. memory/style-reference.md - 风格参考（如有）
✓ 3. stories/*/specification.md - 故事规格
✓ 4. stories/*/creative-plan.md - 创作计划
✓ 5. stories/*/tasks.md - 当前任务
✓ 6. spec/tracking/character-state.json - 角色状态
✓ 7. spec/tracking/relationships.json - 关系网络
✓ 8. spec/tracking/plot-tracker.json - 情节追踪（如有）
✓ 9. spec/tracking/validation-rules.json - 验证规则（如有）

📊 上下文加载状态：✅ 完成

How It Works

Automatic Trigger Timing

User executes /write command
This Skill automatically activates
Mandatorily executes the checklist
Outputs a confirmation report
Only then begins writing

Execution Flow

用户: /write 第10章

         ↓

[pre-write-checklist 自动激活]

         ↓

步骤1：读取 memory/constitution.md
步骤2：读取 memory/style-reference.md（如有）
步骤3：读取 stories/*/specification.md
步骤4：读取 stories/*/creative-plan.md
步骤5：读取 stories/*/tasks.md
步骤6：读取 spec/tracking/character-state.json
步骤7：读取 spec/tracking/relationships.json
步骤8：读取 spec/tracking/plot-tracker.json（如有）
步骤9：读取 spec/tracking/validation-rules.json（如有）

         ↓

输出确认：
📋 写作前检查清单（已完成）：
✓ 1-9 所有文件已读取
📊 上下文加载状态：✅ 完成

关键信息摘要：
- 创作原则：[从constitution提取]
- 当前任务：[从tasks.md提取]
- 主要角色：[从character-state提取]
- 情节进度：[从plot-tracker提取]

         ↓

开始写作第10章...

Output Format

Standard output (all files exist)

📋 写作前检查清单（已完成）：

✓ 1. memory/constitution.md - 创作宪法
   → 核心原则：[列出2-3条关键原则]

✓ 2. memory/style-reference.md - 风格参考
   → 风格要点：[提取关键风格要求]

✓ 3. stories/xxx/specification.md - 故事规格
   → 故事类型：[言情/悬疑/历史等]
   → P0元素：[必须包含的元素]

✓ 4. stories/xxx/creative-plan.md - 创作计划
   → 当前阶段：[第X卷/第X章]
   → 本章目标：[情节/情感目标]

✓ 5. stories/xxx/tasks.md - 当前任务
   → 待写章节：[第X章]
   → 任务状态：[pending/in_progress]

✓ 6. spec/tracking/character-state.json - 角色状态
   → 主要角色：[列出角色名和当前状态]

✓ 7. spec/tracking/relationships.json - 关系网络
   → 核心关系：[主角与谁的关系变化]

✓ 8. spec/tracking/plot-tracker.json - 情节追踪
   → 活跃线索：[当前进行中的情节线]

✓ 9. spec/tracking/validation-rules.json - 验证规则
   → 自动修复：[启用/禁用]

📊 上下文加载状态：✅ 完成（加载9个文件，约XXXX tokens）

🎯 准备写作第X章...

When some files are missing

📋 写作前检查清单（部分完成）：

✓ 1. memory/constitution.md - 创作宪法
✓ 2. ⚠️ memory/style-reference.md - 不存在（可选文件，跳过）
✓ 3. stories/xxx/specification.md - 故事规格
✓ 4. stories/xxx/creative-plan.md - 创作计划
✓ 5. stories/xxx/tasks.md - 当前任务
✓ 6. spec/tracking/character-state.json - 角色状态
✓ 7. spec/tracking/relationships.json - 关系网络
✓ 8. ⚠️ spec/tracking/plot-tracker.json - 不存在（可选文件，跳过）
✓ 9. ⚠️ spec/tracking/validation-rules.json - 不存在（可选文件，跳过）

📊 上下文加载状态：✅ 完成（加载6个必须文件 + 0个可选文件）

💡 建议：运行 `/track-init` 初始化完整追踪系统

When key files are missing (blocks writing)

📋 写作前检查清单（失败）：

✓ 1. memory/constitution.md - 创作宪法
✓ 2. memory/style-reference.md - 风格参考
❌ 3. stories/xxx/specification.md - **文件不存在**
❌ 4. stories/xxx/creative-plan.md - **文件不存在**
❌ 5. stories/xxx/tasks.md - **文件不存在**

⛔ 错误：缺少必需文件，无法继续写作

必须先完成：
1. 运行 `/constitution` 创建创作宪法
2. 运行 `/specify` 定义故事规格
3. 运行 `/plan` 制定创作计划
4. 运行 `/tasks` 分解任务清单

然后才能执行 `/write`

这是seven-step methodology的推荐流程。

Integration with Commands

`/write` command

The checklist must be executed first before writing：

执行顺序：
1. pre-write-checklist（本Skill）→ 读取所有文件
2. 输出确认报告
3. 检查setting-detector → 是否需要激活知识库
4. 开始实际写作

`/analyze` command

Executing the checklist is also recommended during analysis:

分析前先确保context完整：
1. pre-write-checklist → 重新加载所有文件
2. 基于最新状态执行分析

`/track` command

Trigger the checklist after tracking updates:

更新流程：
1. 用户修改tracking文件
2. 运行 `/track` 更新
3. pre-write-checklist → 重新读取验证

File Importance Classification

Required files (missing ones block writing)

1. memory/constitution.md - 创作原则
3. stories/*/specification.md - 故事规格
4. stories/*/creative-plan.md - 创作计划
5. stories/*/tasks.md - 当前任务
6. spec/tracking/character-state.json - 角色状态
7. spec/tracking/relationships.json - 关系网络

Logic: Without these files, the AI does not know:

What principles to follow
What the story is about
Where the writing currently is
Who the characters are and what state they are in

Optional files (warn when missing but allow continuation)

2. memory/style-reference.md - 风格参考
8. spec/tracking/plot-tracker.json - 情节追踪
9. spec/tracking/validation-rules.json - 验证规则

Logic: These files enhance quality, but are not minimum requirements:

style-reference: some users do not use /book-internalize
plot-tracker: simple stories may not need it
validation-rules: non-essential automation

Anti-Loss-of-Focus Mechanism

Problem Scenario

第1章写作：
- AI记得所有设定
- 质量很好

第10章写作：
- 对话已经很长
- AI开始遗忘第1章的设定

第30章写作：
- 完全忘记早期设定
- 角色性格走样
- 情节自相矛盾

Solution Mechanism

每次写作前：
- 强制重读所有核心文件
- 重新加载完整context
- 像写第1章一样对待第30章

结果：
- 第30章质量 ≈ 第1章质量
- 一致性保持
- 不再失焦

Effect Comparison

Comparison Dimension	Without Checklist	With Checklist
Chapters 1-10	✓ Good quality	✓ Good quality
Chapters 11-30	⚠️ Starts becoming unstable	✓ Remains stable
Chapters 31-50	❌ Clearly loses focus	✓ Still stable
Chapters 51+	❌ Severely loses focus	✓ Stable long-term

Configuration Options

Adjust Strictness

Default: Strict mode(recommended)

"使用严格检查清单模式"
→ 缺少必需文件则阻止写作

Loose mode(not recommended)

"使用宽松检查清单模式"
→ 允许跳过部分文件（不推荐，可能失焦）

Custom Checklist Items

If you have additional important files:

"检查清单请额外包含：
- spec/knowledge/worldbuilding/magic-system.md
- spec/knowledge/characters/protagonist-profile.md"

Performance Optimization

Token Consumption

每次写作的额外token成本：

9个文件读取：
- constitution.md：~200 tokens
- specification.md：~500 tokens
- creative-plan.md：~300 tokens
- tasks.md：~150 tokens
- character-state.json：~200 tokens
- relationships.json：~150 tokens
- 其他：~200 tokens

总计：约1700 tokens/次写作

收益：
- 避免失焦导致的重写（节省数万tokens）
- 保持质量一致（用户满意度）
- 长篇项目的可持续性

Extremely high ROI: 1700 tokens in exchange for long-term stable quality.

Caching strategy

同一写作会话中：
第1次写作：读取所有文件（1700 tokens）
第2次写作（1小时内）：检查文件是否修改
- 未修改：使用缓存（0 tokens）
- 已修改：重新读取（部分tokens）

Frequently asked questions

Q: Do I have to read so many files every time I write? Won't it be very slow?

A: No.

File reading is very fast (millisecond level)
Token consumption is reasonable (~1700 tokens)
What you get in return is long-term quality assurance

Comparison：

Without a checklist: Chapter 30 has poor quality → user asks to rewrite 10 chapters → wastes tens of thousands of tokens
With a checklist: +1700 tokens per chapter → even 50 chapters is only +85000 tokens → but quality is stable

Q: Can I skip the checklist?

A: Technically yes, butstrongly not recommended。

"跳过检查清单，直接写作"
→ AI会警告："不推荐，可能导致失焦"
→ 但会尊重你的选择

At your own risk: If you lose focus after chapter 30, don't say I didn't warn you 😊

Q: What if I really don't have some files?

A: There are two cases:

Missing required files(constitution, specification, etc.): → Block writing and prompt to run the corresponding command first to create them

Missing optional files(style-reference, plot-tracker): → Warn but allow continuation, and suggest creating them later

Q: What is the relationship between the checklist and setting-detector?

A: They work complementarily:

pre-write-checklist：
- 加载项目特定文件（你的故事数据）

setting-detector：
- 加载通用知识库（类型惯例、写作技巧）

两者结合 = 完整context：
你的故事设定 + 类型专业知识

Q: Does a 100-chapter novel also need to be read every time?

A: Yes, and it iseven more necessary。

长篇小说的挑战：
- 设定更复杂
- 角色更多
- 情节线更多
- AI更容易忘记

检查清单的作用：
- 确保第100章和第1章质量一致
- 防止角色性格突变
- 防止情节自相矛盾

这是长篇小说质量保证的基石。

Best practices

1. Keep files updated

The checklist can only ensure that the AI reads the files, but the file contents must be accurate:

✓ 角色状态变化 → 更新 character-state.json
✓ 关系变化 → 更新 relationships.json
✓ 新情节线 → 更新 plot-tracker.json

2. Run regularly `/track`

建议频率：每5-10章运行一次 `/track`
作用：
- 更新tracking文件
- 验证一致性
- 发现潜在问题

3. Manually trigger after important changes

如果你手动修改了关键文件：
"请重新执行检查清单，重新加载所有文件"

确保AI看到最新状态。

4. Use together with consistency-checker

pre-write-checklist（写前）：
- 加载所有context
- 准备写作

consistency-checker（写中/写后）：
- 监控一致性
- 发现矛盾

Double assurance = highest quality.

Technical implementation

File reading order

优先级排序（重要的先读）：
1. constitution（最高原则）
2. specification（故事核心）
3. creative-plan（技术方案）
4. tasks（当前任务）
5. character-state（角色数据）
6. relationships（关系数据）
7. plot-tracker（情节追踪）
8. validation-rules（验证规则）
9. style-reference（风格参考）

Error handling

文件不存在：
→ 必需文件：阻止写作，提示创建
→ 可选文件：警告，允许继续

文件格式错误：
→ JSON解析失败：显示错误，建议修复
→ Markdown格式问题：尽力读取，标记问题

文件过大：
→ 超过10000行：警告（可能影响性能）
→ 建议拆分文件

Summary

pre-write-checklist is the Novel Writer Skills v1.0core innovation：

✓ Solves the problem of AI losing focus in long works ✓ Forces rereading of key files ✓ Ensures context completeness ✓ Maintains long-term stable quality ✓ Suitable for professional authors' long-form creation

No longer losing focus after chapter 30 = long-term competitiveness 🎯

This Skill version: v1.0 Last updated: 2025-10-18 Core problem: Solves AI losing focus after chapter 30 Works with: write.md, setting-detector, consistency-checker