worldbuilder

Web Fiction Creation Worldbuilder (worldbuilder)

Trigger Keywords

I want to write a novel, write a web novel, create a novel from scratch, help me write a novel of XX genre, I want to write a novel, give me the whole novel creation process, automatically write a novel, one-stop novel creation service, help me complete a novel, I only have an idea, how do I write a novel, write a web novel from scratch, full-process novel creation

Core Functions

worldbuilder is a one-stop master-control skill for web fiction creation, responsible for coordinating the complete creation chain from the budding of an idea to the finished manuscript, and building a complete and unified story world. Through intelligently orchestrating the invocation order and parameter passing of each specialized skill, it realizes an automated, assembly-line-style web fiction creation experience:

1. Topic Selection Planning: Invoke topic for market analysis and topic positioning, generating the core idea and selling points
2. Outline Design: Invoke outline to build a complete worldview, character settings, volume-by-volume outline, and chapter planning
3. Content Creation: Invoke write to write chapter-by-chapter content according to the outline, maintaining a consistent style
4. Logic Review: Invoke review to verify timeline consistency, character behavior logic, and plot coherence before and after
5. Content Polishing: Invoke polish to optimize writing style, enrich details, and adjust pacing
6. Final Manuscript Verification: Invoke final to complete typo checking, punctuation standardization, final troubleshooting of logic loopholes, and work packaging

Execution Flow

flowchart LR
    A[接收创作需求] --> B[参数校验与初始化]
    B --> C[调用 topic 选题策划]
    C --> D[调用 outline 大纲设计]
    D --> E[循环调用 write 章节撰写]
    E --> F[调用 review 逻辑审查]
    F --> G{审查通过?}
    G -->|否| H[定位问题章节并返工]
    H --> E
    G -->|是| I[调用 polish 内容润色]
    I --> J{润色确认?}
    J -->|否| K[针对性调整润色]
    K --> I
    J -->|是| L[调用 finalize 完稿校验]
    L --> M[生成创作报告与作品文件]

Detailed Process Description：

1. Initialization Phase: Validate the validity of input parameters, create the working directory, and initialize the creation state
2. Requirements Gathering Phase: Intelligently determine whether the information provided by the user is sufficient; if insufficient, automatically trigger interactive questioning to guide the user in supplementing requirements, and after confirming all requirements, proceed to the next phase
3. Topic Selection Phase: Based on the complete gathered requirements, generate 3-5 precisely matching topic options for selection; after confirmation, proceed to outline design
4. Outline Phase: First output the worldview and character settings; after confirmation, generate the complete outline
5. Creation Phase: Create in batches in chapter order, with an interim check every 10 chapters
6. Review Phase: Automatically detect issues such as timeline conflicts, character OOC, and plot contradictions
7. Polishing Phase: Provide multiple polishing style options (concise/detailed/lyrical/passionate, etc.)
8. Final Manuscript Phase: Read from the chapters/ directory the final chapters, output multiple formats (Markdown/HTML/EPUB) to the publish/ directory, and generate a creation summary report

Interactive Requirements Guidance

When the information provided by the user is too brief (only a genre and a few keywords are entered), the system automatically triggers interactive questioning, guiding the user step by step to clarify their creative requirements, ensuring that the generated content fully meets expectations.

Questioning Dimensions (by priority)

Basic Information Confirmation (required)

1. 🎯 Genre Confirmation: Confirm the specific genre subcategory, such as "fantasy" → "high-martial fantasy/cultivation fantasy/otherworld fantasy/system fantasy"
2. 📏 Length Expectation: Confirm the target word count/number of chapters, whether it is a short story/novella/novel/super-long novel
3. 🎯 Core Satisfaction Points: The types of satisfaction points the user values most, such as "face-slapping/leveling up/making money/romance/political intrigue"
4. 👥 Audience Positioning: The target reader group, male-oriented/female-oriented/all-ages, and which age group it leans toward

Core Setting Guidance (optional, depending on the depth of requirements)

1. 🦸 Protagonist Setting Preferences: Protagonist personality (forbearing/flamboyant/scheming/goofy), identity (trash/talent/transmigrator/reincarnator), preference for the type of cheat ability
2. 🎭 Antagonist Setting Preferences: Antagonist type (family enemy/sect rival/foreign race/heavenly dao), antagonist strength
3. 🌍 Worldview Preferences: Which worldview settings are preferred, and whether there are any settings particularly liked/disliked
4. 📖 Reference Works: Whether there are reference works with a similar style, which can match the style more precisely

Style Preference Settings (optional)

1. ✍️ Writing Style: Preference for fast-paced satisfying fiction/refined premium fiction/humorous and funny fiction/dark and oppressive fiction
2. 📱 Publishing Platform: Which platform it is planned to be published on, adapting to the corresponding platform's pacing and word count requirements
3. ⚠️ Taboo Content: Clearly unwanted plots, settings, and character types

Requirement Confirmation Mechanism

• All user answers are automatically saved to .sumeru/session/user-requirements.json, and shared for use across all stages of the entire process
• After the questioning is completed, generate arequirement confirmation summary, and only begin formal creation after the user confirms it is correct
• Modifications during the process are supported; users can adjust previous answers at any time

Data Persistence Specification

All intermediate state data is uniformly stored in the current working directory's .sumeru/ directory, to avoid data loss caused by context compression or cleanup:

Data Lifecycle Management

1. Automatic Saving: Automatically write data to the corresponding directory after each stage is completed, supporting idempotent writes
2. Version Control: Automatically generate version snapshots at key nodes, naming format {stage}-{timestamp}.json
3. Breakpoint Recovery: When using the recovery function, automatically read the corresponding stage data from the .sumeru/ directory
4. Cleanup Rules: Supports cleaning up all intermediate data, retaining the most recent 3 versions by default
5. Data Reuse: Can directly reference other projects' .sumeru/ directory data to achieve worldview/character setting reuse

Detailed Process Description：

1. Initialization Stage: Validate the validity of input parameters, create the working directory, and initialize the creation state
2. Requirement Collection Stage: Intelligently determine whether the information provided by the user is sufficient. If the information is insufficient, automatically trigger interactive questions to guide the user in supplementing the requirements. After all requirements are confirmed, proceed to the next stage.
3. Topic Selection Stage: Based on the complete collected requirements, generate 3-5 precisely matched topic proposals for selection. After confirmation, proceed to outline design.
4. Outline Stage: First output the worldview and character settings. After confirmation, generate the complete outline.
5. Creation Stage: Create in batches in chapter order, with a mid-term review every 10 chapters.
6. Review Stage: Automatically detect issues such as timeline conflicts, character OOC, and plot contradictions.
7. Polishing Stage: Provide a variety of polishing style options (concise/detailed/lyrical/passionate, etc.).
8. Finalization Stage: From chapters/ Read the final chapters from the directory and output them in multiple formats (Markdown/HTML/EPUB) to publish/ directory, and generate a creation summary report.

Interactive Requirement Guidance

When the information provided by the user is too brief (only the genre and a few keywords are entered), the system will automatically trigger interactive questions to guide the user step by step in clarifying their creative requirements, ensuring that the generated content fully meets expectations.

Question Dimensions (by priority)

Basic Information Confirmation (required)

1. 🎯 Genre Confirmation: Confirm the specific genre subtype, such as "fantasy" → "high-martial fantasy/cultivation fantasy/otherworld fantasy/system fantasy"
2. 📏 Length Expectation: Confirm the target word count/number of chapters, whether it is a short story/novella/novel/super-long novel
3. 🎯 Core Thrill Points: The types of thrill points the user values most, such as "face-slapping/leveling up/making money/romance/political intrigue"
4. 👥 Audience Positioning: The target reader group, male-oriented/female-oriented/all-ages, and which age group it leans toward

Core Setting Guidance (optional, depending on the depth of requirements)

1. 🦸 Protagonist Setting Preferences: The protagonist's personality (forbearing/flamboyant/scheming/goofy), identity (trash/talent/transmigrator/reincarnator), and preference for the type of cheat ability
2. 🎭 Antagonist Setting Preferences: Antagonist type (family enemy/sect rival/foreign race/heavenly dao), antagonist strength
3. 🌍 Worldview Preferences: What worldview settings they lean toward, and whether there are settings they especially like/dislike
4. 📖 Reference Works: Whether there are reference works with a similar style, which can match the style more precisely

Style Preference Settings (optional)

1. ✍️ Writing Style: Preference for fast-paced thrillers/delicate premium fiction/humorous and funny fiction/dark and oppressive fiction
2. 📱 Publishing Platform: Which platform it is planned to be published on, to adapt to the corresponding platform's pacing and word count requirements
3. ⚠️ Taboo Content: Clearly specify unwanted plots, settings, and character types

Interaction Mode Parameters

Parameter Name	Description
--interactive	Force full interactive questioning to be enabled; even if the user provides sufficient information, the complete requirement confirmation process will still be followed.
--quick	Quick mode, asking only the 3 most core questions (genre confirmation, length, core thrill points), with defaults used for the rest.
--no-interactive	Turn off interactive questioning and generate directly based on the existing information, suitable for users who clearly know their needs.

Requirement Confirmation Mechanism

• All user answers are automatically saved to .sumeru/session/user-requirements.json, shared across all stages of the workflow
• Generated after the questions are completedRequirements confirmation summary, formal creation starts only after the user confirms it is correct
• Supports mid-process modifications, users can adjust previous answers at any time

Guided examples

> /worldbuilder 玄幻 "废柴逆袭"
🤖 我来帮您完善创作需求，只需要回答几个简单问题：
1️⃣ 请问您想要的玄幻细分类型是？[高武玄幻/修仙玄幻/异世玄幻/系统玄幻/其他]
> 系统玄幻
2️⃣ 预期总篇幅大概多少字？[20万内/20-50万/50-100万/100万以上]
> 100万以上
3️⃣ 您最看重的核心爽点是？[打脸/升级/扮猪吃虎/收小弟/开后宫/其他]
> 打脸+扮猪吃虎
4️⃣ 主角性格偏好？[隐忍腹黑/张扬霸道/逗比搞笑/温柔沉稳/其他]
> 隐忍腹黑
5️⃣ 有没有特别喜欢的参考作品？比如类似《XX》的风格
> 类似《大王饶命》的搞笑风格
...

✅ 需求收集完成，给您确认一下：
类型：系统玄幻
篇幅：100万字以上
核心爽点：打脸+扮猪吃虎
主角性格：隐忍腹黑
风格参考：《大王饶命》搞笑风
是否确认？[Y/n]
> Y
🚀 开始创作！

Parameter description

Parameter name	Type	Required	Default value	Description
genre	string	Yes	-	Work type, such as: fantasy, urban, xianxia, sci-fi, romance, suspense, etc.
keywords	string	Yes	-	Core creative keywords, supports multiple keywords connected with "+", such as: "underdog counterattack + system flow + live-in son-in-law"
--title	string	No	Automatically generated	Specify the work title; if not provided, it will be generated automatically
--length	string	No	"medium"	Expected length: short (within 200,000 words)/medium (200,000-500,000 words)/long (500,000-1,000,000 words)/epic (over 1,000,000 words)
--style	string	No	"balanced"	Writing style: fast (fast-paced)/balanced (balanced)/detailed (detailed)/literary (literary)
--tone	string	No	"neutral"	Overall tone: humorous (humorous)/serious (serious)/inspiring (inspiring)/dark (dark)
--output-dir	string	No	"./output"	Work output directory path
--resume	string	No	-	Interrupt recovery, pass in the creation session ID from last time
--skip-stages	string	No	-	Skip specified stages, comma-separated: topic,outline,write,review,polish,final
--auto-confirm	boolean	No	false	Whether to automatically confirm all intermediate steps (unattended mode)

Usage examples

Basic usage

# 最简单的调用方式，只指定类型和关键词
/worldbuilder 玄幻 "废柴逆袭+系统流"

# 都市言情作品，指定标题
/worldbuilder 言情 "霸道总裁+契约恋爱" --title "总裁的契约新娘"

# 科幻悬疑，长篇幅，快节奏风格
/worldbuilder 科幻 "时间循环+密室解谜" --length epic --style fast

Advanced usage

# 指定详细参数的完整调用
/worldbuilder 仙侠 "重生+无敌流+宗门" \
    --title "重生之太上掌门" \
    --length long \
    --style detailed \
    --tone inspiring \
    --output-dir "./my-novels/taishang" \
    --auto-confirm false

# 从中断点恢复创作
/worldbuilder 玄幻 "废柴逆袭+系统流" --resume "session-20240315-abc123"

# 跳过选题阶段，直接从已有大纲继续创作
/worldbuilder 都市 "职场+重生" --skip-stages topic

Multi-style combinations

# 幽默风都市修仙
/worldbuilder 都市 "修仙+打工+搞笑" --style balanced --tone humorous

# 暗黑系悬疑推理
/worldbuilder 悬疑 "连环杀人+心理侧写+反转" --style detailed --tone dark

# 热血励志竞技
/worldbuilder 竞技 "篮球+天赋+逆袭" --style fast --tone inspiring

Error handling instructions

Common error types and solutions

Error code	Error message	Cause analysis	Solution
INVALID_GENRE	Unsupported work type	The passed-in genre parameter is not in the supported list	Check the type spelling. Supported types: fantasy, urban, xianxia, sci-fi, romance, suspense, historical, game, competitive, military, wuxia, light novel
KEYWORDS_TOO_LONG	Keywords too long	The keywords parameter exceeds the 100-character limit	Simplify the keywords and keep the 3-5 most core ones
DEPENDENCY_MISSING	Missing dependency skills	The required sub-skills (topic, etc.) are not installed	Run /find-skills to find and install all dependency skills
OUTPUT_DIR_PERMISSION	No permission for output directory	The specified output-dir has no write permission	Switch to a directory with permission, or use the default directory
INVALID_SESSION_ID	Invalid session ID	The session ID passed in the resume parameter does not exist	Check whether the session ID is correct, or start creation again
STAGE_SKIP_CONFLICT	Stage skip conflict	The skipped stage has a dependency relationship with subsequent stages	Remove the skip of the preceding stage, or provide the necessary preceding files
CONTENT_GENERATION_FAILED	Content generation failed	Content review or model limitations were encountered during the creation process	Adjust keywords or style parameters, or confirm manually in stages

Error Recovery Mechanism

• Automatic Retry: For temporary network errors, automatically retry 3 times with a 5-second interval
• Checkpoint Saving: Automatically save the state after each stage is completed, supporting recovery from a checkpoint
• Rollback Option: For unsatisfactory stages, you can choose to roll back to the previous node and start over
• Error Reporting: Generate a detailed error log file, located in {output-dir}/error.log

Advanced Usage Scenarios

Scenario 1: Team Collaborative Creation

# 策划完成选题和大纲后，交由写手继续
/worldbuilder 玄幻 "废柴逆袭+系统流" --skip-stages write,review,polish,final

# 写手接手，从创作阶段继续
/worldbuilder 玄幻 "废柴逆袭+系统流" --skip-stages topic,outline --resume "session-xxx"

Scenario 2: Multi-version Comparative Creation

# 生成多个版本进行对比
/worldbuilder 言情 "穿越+宫斗" --title "清宫·甄嬛传" --style literary
/worldbuilder 言情 "穿越+宫斗" --title "清宫·步步惊心" --style fast

Scenario 3: Customized Series Works

# 第一部
/worldbuilder 玄幻 "系统+升级" --title "武帝降临" --length medium

# 第二部（沿用世界观）
/worldbuilder 玄幻 "系统+升级" --skip-stages topic,outline --resume "session-wudi1" --title "武帝降临2"

Scenario 4: A/B Testing Optimization

# 测试不同开篇风格
/worldbuilder 都市 "重生+商战" --skip-stages write,review,polish,final
# 手动修改大纲中的开篇设定后继续
/worldbuilder 都市 "重生+商战" --skip-stages topic --resume "session-xxx"

Scenario 5: Batch Generation of Asset Library

# 生成多个选题方案用于后续选择
/worldbuilder 玄幻 "废柴" --skip-stages outline,write,review,polish,final
/worldbuilder 玄幻 "系统" --skip-stages outline,write,review,polish,final
/worldbuilder 玄幻 "重生" --skip-stages outline,write,review,polish,final

Data Persistence Specifications

All intermediate state data is uniformly stored in the current working directory's .sumeru/ directory to avoid data loss caused by context compression or cleanup:

Global Storage Structure

.sumeru/
├── session/          # 会话全局数据
│   ├── config.json   # 创作配置参数
│   ├── status.json   # 当前进度状态
│   └── history.log   # 操作历史记录
├── topic/            # 选题阶段输出
│   ├── report.md     # 完整选题策划报告
│   └── options.json  # 多选题方案原始数据
├── outline/          # 大纲阶段输出
│   ├── world.md      # 世界观设定
│   ├── characters.json # 人物设定卡
│   ├── plot.md       # 剧情大纲
│   └── chapters.json # 章节细纲
├── write/            # 创作阶段输出
│   ├── draft/        # 章节草稿
│   └── progress.json # 创作进度跟踪
├── review/           # 审查阶段输出
│   ├── timeline.json # 时间线数据
│   ├── issues.json   # 问题清单
│   └── plot-map.json # 剧情脉络图
├── polish/           # 润色阶段输出
│   ├── modified/     # 润色后版本
│   └── diff.json     # 修改对比记录
└── finalize/         # 完稿阶段输出
    ├── clean/        # 纯净版全文
    ├── platforms/    # 各平台导出版本
    └── report.md     # 完稿校验报告

Data Lifecycle Management

1. Automatic Saving: Automatically write data to the corresponding directory after each stage is completed, supporting idempotent writes
2. Version Control: Automatically generate version snapshots at key nodes, naming format {stage}-{timestamp}.json
3. Checkpoint Recovery: When using the --resume parameter, automatically read the corresponding stage data from the .sumeru/ directory
4. Cleanup Rules: Supports the --clean parameter to clean up all intermediate data, keeping the most recent 3 versions by default
5. Data Reuse: Can directly reference the .sumeru/ directory data of other projects to achieve worldbuilding/character setting reuse

Advanced Configuration: Custom Stage Hooks

Through the configuration file {output-dir}/hooks.json , custom processing can be inserted before and after each stage:

{
  "before_topic": "my-preprocess-script.sh",
  "after_outline": "validate-outline.js",
  "before_write": "setup-write-env.py",
  "after_final": "deploy-to-platform.sh"
}