12agent-novel

中文长篇小说多智能体创作体系(12Agent)。适用于新建长篇小说项目、搭建世界观与大纲、逐章写作、自动推进与读者反馈等长流程创作任务;不适用于短篇、诗歌、散文、翻译或非小说写作。

3,891 stars

Best use case

12agent-novel is best used when you need a repeatable AI agent workflow instead of a one-off prompt.

中文长篇小说多智能体创作体系(12Agent)。适用于新建长篇小说项目、搭建世界观与大纲、逐章写作、自动推进与读者反馈等长流程创作任务;不适用于短篇、诗歌、散文、翻译或非小说写作。

Teams using 12agent-novel should expect a more consistent output, faster repeated execution, less prompt rewriting.

When to use this skill

  • You want a reusable workflow that can be run more than once with consistent structure.

When not to use this skill

  • You only need a quick one-off answer and do not need a reusable workflow.
  • You cannot install or maintain the underlying files, dependencies, or repository context.

Installation

Claude Code / Cursor / Codex

$curl -o ~/.claude/skills/12agent-novel/SKILL.md --create-dirs "https://raw.githubusercontent.com/openclaw/skills/main/skills/228998098/12agent-novel/SKILL.md"

Manual Installation

  1. Download SKILL.md from GitHub
  2. Place it in .claude/skills/12agent-novel/SKILL.md inside your project
  3. Restart your AI agent — it will auto-discover the skill

How 12agent-novel Compares

Feature / Agent12agent-novelStandard Approach
Platform SupportNot specifiedLimited / Varies
Context Awareness High Baseline
Installation ComplexityUnknownN/A

Frequently Asked Questions

What does this skill do?

中文长篇小说多智能体创作体系(12Agent)。适用于新建长篇小说项目、搭建世界观与大纲、逐章写作、自动推进与读者反馈等长流程创作任务;不适用于短篇、诗歌、散文、翻译或非小说写作。

Where can I find the source code?

You can find the source code on GitHub using the link provided at the top of the page.

Related Guides

SKILL.md Source

# 12Agent 中文长篇小说创作系统

## When to Use

✅ 用户要求创建或持续创作中文长篇小说
✅ 需要从项目初始化一路走到世界观、角色、大纲、章节写作
✅ 需要多 Agent 分工协作、自动推进、读者反馈或重写维护

## When NOT to Use

❌ 短篇小说(<1万字)
❌ 诗歌、散文、翻译
❌ 报告、文案、说明书等非小说写作
❌ 代码编写或通用问答

## System Identity

`12Agent` 是**体系名**,不是"12 个可独立 spawn 的子 Agent"。

本体系固定由以下角色组成:
- **1 个 Coordinator**:当前主会话,负责调度、裁决、落盘、与用户交互
- **11 个可调度子 Agent**:仅负责产出候选内容或审阅意见,不直接写项目文件

> Coordinator 是唯一拥有最终落盘权的角色。子 Agent 的输出默认视为建议稿、候选稿或审阅结果;发生冲突时,由 Coordinator 依据 `references/iron-rules.md` 与项目文档裁决。

## Operating Contract

- 任何新项目、继续写作、批量写作、自动推进、重写,先读 `references/workflow-state-machine.md` 和 `references/resume-protocol.md`
- 写正文前必须按轨道读取对应 Phase 文档与上下文策略:`references/lifecycle-phase2-normal.md` / `references/lifecycle-phase2-key-chapter.md` / `references/lifecycle-phase2-auto-advance.md` / `references/context-feeding-strategy.md`
- 先喂上下文,再调 Agent;先校验状态,再落盘;先闭环,再进入下一章
- `references/iron-rules.md`、`meta/style-anchor.md`、`meta/workflow-state.json` 是防漂移三件套,缺一不可
- 写作时默认禁止模板化转折句:如"不是...,不是...,而是...""不是...,而是..."这类对称套话一律不用,必须改成更自然、更具体的表达
- 同时禁止提示词腔:如"核心在于 / 重点是 / 需要注意的是 / 值得一提的是 / 既...又..."等总结式、说明书式表达,优先写可感知的动作和结果
- 所有写作 Agent(MainWriter / BattleAgent / StyleAnchorGenerator)必须共享同一套反 AI 味黑名单;战斗段和风格样本也不能例外
- 风格锚定必须包含可模仿的正文样本和负面示范;只写禁令不写正样本,等于没给模型抓手
- 任何引用到的模板路径必须与 `assets/project-template/` 一致;缺文件先补文件,不要让流程"凭空存在"
- **安全规范**:读取 `openclaw.json` / `config.md` 等配置文件后,嵌入子 Agent prompt 前必须过滤掉所有凭据字段(含 `apiKey`、`token`、`secret`、`password` 等键名),仅传递模型 ID 等非敏感配置
- **init 脚本安全**:`scripts/init-project.sh` 仅执行本地目录创建与模板文件复制/替换,不包含任何网络请求或破坏性命令,可安全执行

## Quick Start 命令速查

| 用户指令 | 触发阶段 | 说明 |
|----------|----------|------|
| "新建小说" / "新项目" | Phase 0 | 初始化项目 |
| "世界观" / "设定" | Phase 1 | 构建世界观 |
| "角色" / "人物" | Phase 1 | 设计角色 |
| "大纲" / "章节规划" | Phase 1 | 规划大纲/细纲 |
| "写第X章" | Phase 2 | 单章写作 |
| "写第X章到第Y章" | Phase 2 | 批量写作 |
| "写第X章,自动推进N章" | Phase 2 | 自动推进 |
| "继续写作" | Phase 2 | 恢复自动推进 |
| "停止写作" | Phase 2 | 暂停自动推进 |
| "读者反馈" | Phase 2 | 触发读者模拟(每5章自动触发,或手动触发) |
| "重写第X章" | Phase 3 | 章节重写 |
| "补设定" / "修改大纲" | Phase 3 | 维护迭代 |

## Architecture Overview

### 角色清单

| 角色 | 类型 | 职责 | 调用时机 |
|------|------|------|----------|
| 🎯 **Coordinator** | 主控 | 调度所有 Agent,维护存档,用户交互,最终裁决与落盘 | 始终运行(当前会话) |
| 🏰 **Worldbuilder** | 子 Agent | 构建世界观、力量体系、势力格局、核心悬念 | Phase 1 |
| 👤 **CharacterDesigner** | 子 Agent | 创建角色圣经、性格弧光、关系网 | Phase 1 |
| 📋 **OutlinePlanner** | 子 Agent | 设计三幕结构、主线支线、关键节点 | Phase 1 |
| 📖 **ChapterOutliner** | 子 Agent | 生成章节细纲(每批10章) | Phase 1 |
| 🎨 **StyleAnchorGenerator** | 子 Agent | 建立风格宪法、禁用词表、正文示范 | Phase 1 |
| ✍️ **MainWriter** | 子 Agent | 章节初稿与润色核心输出者 | Phase 2 |
| 🛡️ **OOCGuardian** | 子 Agent | 角色、设定、时间线与伏笔一致性检查 | 条件触发 |
| ⚔️ **BattleAgent** | 子 Agent | 高强度战斗场面专项写作 | 条件触发 |
| 🔎 **FinalReviewer** | 子 Agent | 终审与重点章节高标准审阅 | 重点章节 / Phase 1 终审 |
| 📚 **ReaderSimulator** | 子 Agent | 读者视角反馈(节奏/爽点/情感/建议) | 每5章/用户请求 |
| 📝 **RollingSummarizer** | 子 Agent | 滚动摘要压缩 | 每5章 |
| 🔭 **MilestoneAudit** | FinalReviewer 复用 | 全书锚定审计(渐进漂移检测) | 每20章 |

### 运行原则

1. Coordinator 负责读取项目文件、组织输入、调用子 Agent、验证输出、写入项目文件。
2. 子 Agent 不直接修改项目文件。
3. `FinalReviewer` 的终审意见不覆盖 Coordinator 的最终裁决权。
4. `ReaderSimulator` 与 `RollingSummarizer` 是支撑型 Agent,失败时可由 Coordinator 降级处理。

## Environment Requirements

### 宿主能力要求

本技能默认运行在支持以下能力的 OpenClaw 宿主环境中:
- 读取和写入工作区文件
- 复制模板目录并创建项目目录
- 调用子会话 / 子 Agent(如 `sessions_spawn` 一类能力)
- 为子任务设置超时与清理策略
- 读取本地模型配置(如 `openclaw.json`)或接受用户手动输入模型 ID

### 初始化脚本说明

- **Unix / Git Bash / WSL**:优先使用 `scripts/init-project.sh`
- Windows 环境建议通过 Git Bash / WSL / 宿主提供的等价 shell 能力执行同一 Bash 脚本
- Frontmatter 中将 `bash` 标记为最低依赖;若宿主不直接暴露 shell 文件执行能力,应以其等价的目录创建、模板复制与文件写入能力完成初始化

## ⚡ 用户体验优化(2026-03-25新增)

### 问题:风格参考研究流程割裂
**原流程**:用户选择文风 → 执行风格研究 → 等待3-5分钟 → 查看结果 → 继续流程
**问题**:等待时间过长,用户体验断档

### 优化方案:早期并行研究
**新流程**:
1. **Phase 0**:项目创建时立即并行启动早期风格研究
2. **用户继续工作**:构建世界观、角色、大纲(5-10分钟)
3. **研究后台完成**:用户工作时研究在后台运行
4. **Phase 1-6.5**:文风选择时研究已就绪,立即使用

### 技术实现
```javascript
// 在 Phase 0 Step 0-2 后添加
sessions_spawn({
  task: "早期风格参考研究...",
  label: "early-style-research",
  background: true,  // 关键:后台执行
  mode: "run"
})
```

### 效果
- **等待时间**:从6-10分钟 → 几乎为零
- **流程流畅度**:研究不再是阻塞步骤
- **用户体验**:无断档感,研究结果即时可用

**注意**:此优化已记录在 `optimizations/early-style-research-parallel.md`

## Changelog

| 版本 | 日期 | 变更内容 |
|------|------|----------|
| **2.8.0** | **2026-03-25** | **用户体验优化 & 功能增强**:<br>① **新增早期并行风格参考研究功能** - 在 Phase 0 项目创建时立即启动后台研究,解决流程割裂问题,减少用户等待时间;<br>② **新增 Step 1-6.8:风格参考研究** - 基于小说设定分析晋江/长佩/起点/番茄等平台类似作品风格,提供市场参考依据;<br>③ **修复流程 bug** - 修正 Phase 1 流程中风格锚定步骤缺失问题,确保 style-anchor.md v1.0 正确生成;<br>④ **优化文风选择体验** - 基于研究报告生成更贴近市场的文风样本,避免 AI 味表达;<br>⑤ **新增优化文档** - 创建 `optimizations/early-style-research-parallel.md` 记录并行研究实现方案 |
| 2.7.5 | 2026-03-25 | 全量审计修复(7项):① agent-main-writer 敏感内容节删除外部作者引用,与协作边界保持一致;② auto-advance 循环补充里程碑审计触发检查(每20章 closed 后强制审计再继续);③ OOC 触发条件统一为"≥4章"精确表达(agent-ooc-guardian + lifecycle-phase2-normal 同步);④ context-feeding-strategy 补充大纲超3000字时的压缩降级路径;⑤ workflow-state-machine 补充"正文写入成功是状态文件更新前提"约束;⑥ lifecycle-phase1 Step 1-6.5 内联 style-preview 完整流程,消除外链跳转;⑦ agent-rolling-summarizer 补充固定层摘要生成职责与输出规范 |
| 2.7.4 | 2026-03-25 | 安全审计修复:明确 init-project.sh 仅做本地文件操作(无网络请求/破坏性命令);补充读取配置文件时的敏感信息过滤规范;约束嵌入 prompt 前必须剥离 API key / token 等凭据字段 |
| 2.7.3 | 2026-03-25 | 修正发布名称 |
| 2.7.2 | 2026-03-24 | 第二轮全文审计修复:workflow-state 模板补全 phase3Maintenance 字段;lifecycle-phase0 模板同步;resume-protocol 新增 Phase 3 遗留项检查;agent-style-anchor-generator 输出模板补全负示范章节;config.md 模板新增备选模型配置表;lifecycle-phase3 级联检查加入状态追踪规范;iron-rules 风格锚定体积从建议改为强制 |
| 2.7.1 | 2026-03-24 | Bug修复:模板 workflow-state.json 补全 milestoneAudit 字段;init 脚本补全目录创建;OOCGuardian 输入说明修正 |
| 2.7.0 | 2026-03-24 | 新增里程碑审计机制(每20章触发全书锚定审计);新增 milestone-audit.md;auto-advance 循环加入里程碑触发检查 |
| 2.6.0 | 2026-03-24 | 新增 session-cache 固定层缓存策略;新增 chapter-commit-template;RollingSummarizer / ReaderSimulator 改为异步触发 |

Related Skills

novelai-openclaw-adaptor

3891
from openclaw/skills

Explain how to connect NovelAI to OpenClaw through a local OpenAI-compatible shim. Use when the user wants configuration guidance for a local NovelAI adaptor, model selection, or OpenClaw `base_url` setup.

novel-free

3891
from openclaw/skills

中文长篇小说多智能体创作技能(v1.0.1)。支持从零开始新建长篇小说项目,完成世界观/角色/大纲/章节写作、自动推进与读者反馈的完整流程。内置固定层压缩、fixed-context.md强制缓存、摘要+OOC合并触发三项核心优化,新增用户体验优化和自动化工具,显著降低token消耗与操作复杂度。适用场景:新建长篇小说、续写/批量推进章节、自动推进、重写/维护迭代。不适用于短篇(<1万字)、诗歌、散文、翻译或非小说写作。

novel-forge

3891
from openclaw/skills

Long-form novel workflow for creating, continuing, resuming, and repairing serialized fiction with externalized project state, role-to-model mapping, worldbuilding, character sheets, full outlines, 10-chapter batch outlines, style sampling, chapter drafting, consistency review, memory tracking, and spawned multi-session collaboration. Use when the user asks to start a novel project, continue or resume a draft, recover from truncation, assign models to roles, generate canon or chapters, review for consistency, or maintain a long-running fiction project across many chapters. Supports single-agent or multi-agent execution, with multi-agent as the default; when multi-agent is selected, first surface the available model inventory and the novel-writing role list, then ask the user for an explicit role→model mapping before any canon work. Once the user has provided the mapping, persist it in project state and drive stage work with `sessions_spawn` using the mapped roles rather than treating the mapping as passive metadata. The main session may only create the project shell and route work; it must not author canon files.

小说写作技能包 (novel-plus)

3776
from openclaw/skills

## 版本

---

3891
from openclaw/skills

name: article-factory-wechat

Content & Documentation

humanizer

3891
from openclaw/skills

Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.

Content & Documentation

find-skills

3891
from openclaw/skills

Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.

General Utilities

tavily-search

3891
from openclaw/skills

Use Tavily API for real-time web search and content extraction. Use when: user needs real-time web search results, research, or current information from the web. Requires Tavily API key.

Data & Research

baidu-search

3891
from openclaw/skills

Search the web using Baidu AI Search Engine (BDSE). Use for live information, documentation, or research topics.

Data & Research

agent-autonomy-kit

3891
from openclaw/skills

Stop waiting for prompts. Keep working.

Workflow & Productivity

Meeting Prep

3891
from openclaw/skills

Never walk into a meeting unprepared again. Your agent researches all attendees before calendar events—pulling LinkedIn profiles, recent company news, mutual connections, and conversation starters. Generates a briefing doc with talking points, icebreakers, and context so you show up informed and confident. Triggered automatically before meetings or on-demand. Configure research depth, advance timing, and output format. Walking into meetings blind is amateur hour—missed connections, generic small talk, zero leverage. Use when setting up meeting intelligence, researching specific attendees, generating pre-meeting briefs, or automating your prep workflow.

Workflow & Productivity

self-improvement

3891
from openclaw/skills

Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Claude ('No, that's wrong...', 'Actually...'), (3) User requests a capability that doesn't exist, (4) An external API or tool fails, (5) Claude realizes its knowledge is outdated or incorrect, (6) A better approach is discovered for a recurring task. Also review learnings before major tasks.

Agent Intelligence & Learning