sumeru-write

Web Novel Chapter Writing

Trigger Keywords

Core Features

1. Generation Based on Complete Detailed Outline: Automatically reads

   .sumeru/outline/chapter-outlines.json

   and generates chapters in batches according to the detailed outline
2. Intelligent Detailed Outline Matching: Supports generation by chapter number, volume number, or all chapters
3. Generate complete chapter content based on outline and detailed outline
4. Automatically adapt to web novel rhythm: eye-catching opening, conflicts in the middle, suspenseful ending
5. Maintain consistency of character personalities and plot logic
6. Support custom chapter length (default 4000-5000 words per chapter)
7. Support continuation, modification, and adjustment of existing chapter content

Continuation Rules

Continuation Modes

Continuation Notes

1. Maintain consistency of character personalities, no OOC (Out Of Character)
2. Maintain the power system and worldview set in the previous content without collapse
3. Natural foreshadowing resolution, no abruptness
4. Keep language style consistent with the previous content 5.承接上文剧情，开启下文伏笔（注：此处原文逻辑承接，翻译为"Connect with the previous plot and lay foreshadowing for the next part"）
5. If the existing chapter content is incomplete, prioritize completing it

Multi-Chapter Generation

Outline-Driven Batch Generation

Automatic Outline Reading Mode

.sumeru/outline/chapter-outlines.json

# Generate all chapters (read outline, automatic parallel processing)
/sumeru-write 全部章节

# Generate chapters in specified range
/sumeru-write 第1-50章

# Generate all chapters in specified volume
/sumeru-write 第1卷

# Generate specific chapters
/sumeru-write 第3章,第5章,第10章

# Batch parallel creation for specified range
/sumeru-write 第1-100章 批量并行

Supported Outline Input Formats

# Generate single chapter using natural language description
/sumeru-write 第3章 "主角在拍卖会上获得神秘功法"

# Generate based on existing detailed outline
/sumeru-write 第3章 按细纲生成

Detailed Outline Data Structure Verification

• Check if required fields exist
• Verify if character names are defined in characters.json
• Check if scene locations are defined in world.json
• Provide suggestions for supplementing missing information

Sub-Agent Parallel Batch Writing (Recommended for Large Number of Chapters)

• Required number of Agents = ceil(total chapters / 3), automatically calculated by the scheduler
• Adjacent chapters are assigned to the same Agent to maintain context coherence

• ✅ Outline Isolation: Each sub-agent only obtains the detailed outline of the chapters it is responsible for, avoiding context overflow
• ✅ 3-Chapter Limit Guarantee: Each Agent handles a maximum of 3 chapters to ensure generation quality and consistency
• ✅ Context Isolation: Each sub-agent does not carry historical chapter content, completely solving the problem of long context compression/overflow
• ✅ Speed Improvement: Multi-parallel writing, speed is N times that of serial writing
• ✅ Error Isolation: Failure of single-chapter generation does not affect other chapters, automatically retry failed chapters
• ✅ Memory Optimization: Sub-agents are automatically destroyed after completion to release memory resources
• ✅ Incremental Writing: Save to

  .sumeru/write/draft/

  immediately after finishing each chapter, no need to wait for all to complete
• ✅ Progress Visualization: Real-time display of completed/in-progress/pending chapter status

flowchart LR
    A[批量写作任务] --> B[读取chapter-outlines.json]
    B --> C[验证细纲完整性]
    C --> D[创建任务队列，按每Agent最多3章分配]
    D --> E[计算所需Agent数 = ceil/总章数/3/]
    E --> F[启动N个并行子agent]
    F --> G[子Agent拉取任务 → 获取对应章节细纲 → 生成章节 → 保存文件]
    G --> H{队列是否为空?}
    H -->|否| G
    H -->|是| I[汇总进度，生成完成报告]

• Allocate continuously in chapter order, e.g., Agent1 is responsible for chapters 1-3, Agent2 for chapters 4-6, and so on
• Agents with less than 3 remaining chapters at the end are allocated according to the actual number of remaining chapters
• Adjacent chapters are assigned to the same Agent to maintain context coherence

• Complete chapter content, conforming to the specified style and rhythm
• Preview/idea suggestions for the next chapter content
• Sorting of key plot points of this chapter
• Foreshadowing hints embedded in this chapter (optional)
• Summary of character growth/changes (optional)

Chapter File Naming Specification

{three-digit chapter number}-{chapter title}.md

001

003

042

128

-

-

第一次解析

废物觉醒系统

.md

.md

003-第一次解析.md
001-废物觉醒系统.md
042-时间线冲突.md
128-终极决战.md

第3章-第一次解析.md    # Do not add "Chapter X" prefix
3-第一次解析.md         # Chapter number must be three-digit zero-padding
003_第一次解析.md       # Separator is hyphen, not underscore
003-第一次解析.txt      # Extension must be .md
003第一次解析.md        # Must have hyphen separator

Data Persistence

Official Output (Visible to Users)

• Generated chapters are saved by default in the

  chapters/

  directory of the current working directory, strictly following the above naming specification
• Chapter files are pure text content without any intermediate marks or metadata, which can be directly read and edited by users
• Support custom chapter output directory
• Batch Generation Progress Report:

  chapters/WRITE_PROGRESS.md

  (updated in real-time)

Intermediate Process Data (Internal Use Only)

.sumeru/write/

• progress.json

  : Creation progress tracking, including completed chapters, word count statistics, and status of each chapter

• chapter-meta.json

  : Metadata for each chapter, including core events, appearing characters, highlight positions, foreshadowing records

• character-state.json

  : Dynamic tracking of character states, recording character abilities, relationships, and state changes at various time points

• used-outlines.json

  : Records of used chapter detailed outlines, supporting incremental generation

• original/

  : Backup directory for original chapter files. When reviewing or polishing and modifying files in

  chapters/

  , the original chapter files will be automatically backed up to the

  .sumeru/write/original/

  directory first to ensure rollback capability

Cooperation with Other Skills

• Predecessor Skills: Automatically reads outline data from the

  .sumeru/outline/

  directory
  • Use

    characters.json

    to maintain character personality consistency
  • Use complete chapter detailed outlines in

    chapter-outlines.json

    to drive batch generation
  • Use

    world.json

    to maintain worldview setting consistency
• Successor Skills: Generated chapter data can be used by

  sumeru-review

  ,

  sumeru-polish

  ,

  sumeru-finalize

Breakpoint Recovery

• Read existing chapter files in the

  chapters/

  directory and progress from

  .sumeru/write/progress.json

  each time the task starts
• Read the complete detailed outline from

  .sumeru/outline/chapter-outlines.json

• Continue from the latest unfinished chapter, automatically skip generated chapters
• Support resuming creation from a specified chapter
• Support generating only missing chapters (incremental mode)