docs: update README for v0.4 — spinoff writing, style cloning, post-write validator

Bump version to 0.4.0. Add v0.4 update section (zh + en) covering:
- Spinoff writing with canon import and 4 audit dimensions
- Style cloning from real novels (statistical + LLM guide)
- Post-write validator (11 deterministic rules + auto spot-fix)
- Audit-revise loop hardening (spot-fix default, AI marker guard, temp lock)
- Dimension count 26→32, scheduler rewrite, web search for auditor

Update roadmap, command table, project structure, and revise mode list.

README.en.md

@@ -20,6 +20,75 @@
 
 Open-source multi-agent system that autonomously writes, audits, and revises novels — with human review gates that keep you in control.
 
+## v0.4 Update
+
+Spinoff writing + style cloning + post-write validator + audit-revise hardening.
+
+### Spinoff Writing
+
+Create prequels, sequels, side-stories, or what-if branches from existing books. Spinoffs share the parent's world and characters but have independent plot lines.
+
+```bash
+inkos import canon my-prequel --from main-novel   # Import parent canon
+inkos write next my-prequel                        # Writer auto-reads canon constraints
+```
+
+Generates `story/parent_canon.md` containing the parent's world rules, character snapshots (with information boundaries), key event timeline, and foreshadowing state. The auditor auto-activates 4 spinoff-specific dimensions:
+
+| Dimension | Checks |
+|-----------|--------|
+| Canon Event Conflict | Whether spinoff events contradict the parent's canon constraints |
+| Future Info Leak | Whether characters reference information revealed after the divergence point |
+| Cross-Book World Consistency | Whether the spinoff violates parent world rules (power systems, geography, factions) |
+| Spinoff Foreshadowing Isolation | Whether the spinoff oversteps by resolving parent foreshadowing |
+
+Auto-activates when `parent_canon.md` is detected. No extra configuration needed.
+
+### Style Cloning
+
+Feed in a real novel excerpt. The system extracts a statistical fingerprint + generates a qualitative style guide, then injects both into every chapter's Writer prompt.
+
+```bash
+inkos style analyze reference.txt                      # Analyze: sentence length, TTR, rhetorical features
+inkos style import reference.txt my-book --name "Author"  # Import style into book
+```
+
+Produces two files:
+- `style_profile.json` — statistical fingerprint (sentence/paragraph length distribution, vocabulary diversity, rhetorical density)
+- `style_guide.md` — LLM-generated qualitative guide (rhythm, tone, word preferences, taboos)
+
+The Writer reads the style guide every chapter; the Auditor cross-checks against it in the style dimension.
+
+### Post-Write Validator
+
+11 deterministic rules, zero LLM cost, fires immediately after each chapter:
+
+- Banned sentence patterns, dash prohibition, transition word density limits
+- High-fatigue word caps, meta-narration detection, report terminology in prose
+- Author sermonizing, collective shock cliches, consecutive "le" (了) character runs
+- Paragraph length checks (mobile-reading friendly), book-level prohibitions
+
+When the validator finds error-level violations, it auto-triggers `spot-fix` mode for targeted repair before the LLM audit even runs.
+
+### Audit-Revise Loop Hardening
+
+Real testing showed `rewrite` mode introduces 6x more AI markers than the original text. Now:
+
+- Auto-revise mode changed from `rewrite` to `spot-fix` (only fixes flagged sentences, leaves everything else untouched)
+- Post-revise AI marker guard: if revision increases AI tell count, the revision is discarded
+- Re-audit temperature locked to 0 (deterministic pass/fail gating)
+- `polish` mode boundaries strengthened (no adding paragraphs, renaming entities, or changing causality)
+
+### Other v0.4 Changes
+
+- Audit dimensions expanded from 26 to 32 (+4 spinoff dims + sensitive word check + reader expectation management)
+- Auditor web search: era-research genres can call DuckDuckGo to verify real events/people/geography
+- Scheduler rewrite: AI-paced (15min cycles), parallel book processing, immediate retry, daily cap
+- New `spot-fix` revise mode (targeted repair)
+- `additionalAuditDimensions` in `book_rules.md` now supports name-string matching
+
+---
+
 ## v0.3 Update
 
 Three-layer rule separation + cross-chapter memory + AIGC detection + Webhook.
@@ -112,13 +181,13 @@ fatigueWordsOverride: ["pupils constricted", "disbelief"]   # Override genre def
 
 Protagonist personality lock, numerical caps, custom prohibitions, fatigue word overrides — each book's rules are independent, without affecting the genre template.
 
-### 26-Dimension Audit
+### 32-Dimension Audit
 
-Auditing is broken down into 26 dimensions, with genre-appropriate subsets auto-enabled:
+Auditing is broken down into 32 dimensions, with genre-appropriate subsets auto-enabled:
 
 OOC check, timeline, setting conflicts, foreshadowing, pacing, writing style, information leaking, vocabulary fatigue, broken interest chains, side character intelligence drops, side character tool-ification, hollow payoffs, dialogue authenticity, padding detection, knowledge base contamination, POV consistency, power scaling collapse, numerical verification, era research, paragraph uniformity, hedge word density, formulaic transitions, list-like structure, subplot stagnation, flat emotional arc, monotonous pacing
 
-Xuanhuan/Xianxia: all 26 dimensions. Urban: 24 dimensions (including era research). Horror: 22 dimensions. Dims 20-23 (AI-tell detection) use a pure rule engine — no LLM calls consumed.
+Dims 20-23 (AI-tell detection) + dim 27 (sensitive words) use a pure rule engine — no LLM calls. Dims 28-31 (spinoff) auto-activate when `parent_canon.md` is detected. Dim 32 (reader expectation management) is always on.
 
 ### De-AI-ification
 
@@ -135,7 +204,7 @@ Vocabulary fatigue audit + AI-tell audit (dims 20-23) provide dual detection. St
 ### Other
 
 - Supports OpenAI + Anthropic native + all OpenAI-compatible endpoints
-- Reviser supports polish / rewrite / rework / anti-detect modes
+- Reviser supports polish / rewrite / rework / anti-detect / spot-fix modes
 - Genres without numerical systems skip resource ledger generation
 - All commands support `--json` structured output for OpenClaw / external agent integration
 - Auto-detect book-id when project has only one book
@@ -317,6 +386,7 @@ inkos up                          # Daemon mode
 | `inkos detect [id] [n]` | AIGC detection (`--all` for all chapters, `--stats` for statistics) |
 | `inkos style analyze <file>` | Analyze reference text to extract style fingerprint |
 | `inkos style import <file> [id]` | Import style fingerprint into a book |
+| `inkos import canon [id] --from <parent>` | Import parent canon for spinoff writing |
 | `inkos update` | Update to latest version |
 | `inkos up / down` | Start/stop daemon |
 
@@ -385,8 +455,12 @@ TypeScript monorepo managed with pnpm workspaces.
 - [x] Pluggable radar (RadarSource interface)
 - [x] External agent integration (OpenClaw, etc.)
 - [x] Genre customization + per-book rules (genre CLI + book_rules.md)
-- [x] 26-dimension continuity audit (including AI-tell detection)
+- [x] 32-dimension continuity audit (including AI-tell detection + spinoff dims)
 - [x] De-AI-ification rules + style fingerprint injection
+- [x] Spinoff writing (canon import + 4 audit dimensions + info boundary control)
+- [x] Style cloning (statistical fingerprint + LLM style guide + Writer injection)
+- [x] Post-write validator (11 hard rules + auto spot-fix)
+- [x] Audit-revise loop hardening (AI marker guard + temperature lock)
 - [x] Multi-LLM provider (OpenAI + Anthropic + compatible endpoints)
 - [x] AIGC detection + anti-detect rewrite pipeline
 - [x] Webhook notifications + smart scheduler (quality gates)

README.md

@@ -20,6 +20,85 @@
 
 Agent 写小说。写、审、改，全程接管。
 
+## v0.4 更新
+
+番外写作 + 文风仿写 + 写后验证器 + 审计闭环加固。
+
+### 番外写作（Spinoff）
+
+基于已有书创建前传、后传、外传或 if 线。番外和正传共享世界观和角色，但有独立剧情线。
+
+```bash
+inkos import canon 烈焰前传 --from 吞天魔帝   # 导入正传正典到番外
+inkos write next 烈焰前传                     # 写手自动读取正典约束
+```
+
+导入后生成 `story/parent_canon.md`，包含正传的世界规则、角色快照（含信息边界）、关键事件时间线、伏笔状态。写手在动笔前参照正典，审计员自动激活 4 个番外专属维度：
+
+| 维度 | 审查内容 |
+|------|----------|
+| 正传事件冲突 | 番外事件是否与正典约束表矛盾 |
+| 未来信息泄露 | 角色是否引用了分歧点之后才揭示的信息 |
+| 世界规则跨书一致性 | 番外是否违反正传世界规则（力量体系、地理、阵营） |
+| 番外伏笔隔离 | 番外是否越权回收正传伏笔 |
+
+检测到 `parent_canon.md` 自动激活，无需额外配置。
+
+### 文风仿写
+
+喂入真人小说片段，系统提取统计指纹 + 生成风格指南，后续每章自动注入写手 prompt。
+
+```bash
+inkos style analyze 参考小说.txt                     # 分析：句长、TTR、修辞特征
+inkos style import 参考小说.txt 吞天魔帝 --name 某作者  # 导入文风到书
+```
+
+产出两个文件：
+- `style_profile.json` — 统计指纹（句长分布、段落长度、词汇多样性、修辞密度）
+- `style_guide.md` — LLM 生成的定性风格指南（节奏、语气、用词偏好、禁忌）
+
+写手每章读取风格指南，审计员在文风维度对照检查。
+
+### 写后验证器
+
+11 条确定性规则，零 LLM 成本，每章写完立刻触发：
+
+| 规则 | 说明 |
+|------|------|
+| 禁止句式 | 「不是……而是……」 |
+| 禁止破折号 | 「——」 |
+| 转折词密度 | 仿佛/忽然/竟然等，每 3000 字 ≤ 1 次 |
+| 高疲劳词 | 题材疲劳词单章每词 ≤ 1 次 |
+| 元叙事 | 编剧旁白式表述 |
+| 报告术语 | 分析框架术语不入正文 |
+| 作者说教 | 显然/不言而喻等 |
+| 集体反应 | 「全场震惊」类套话 |
+| 连续了字 | ≥ 4 句连续含「了」 |
+| 段落过长 | ≥ 2 个段落超 300 字 |
+| 本书禁忌 | book_rules.md 中的禁令 |
+
+验证器发现 error 级违规时，自动触发 `spot-fix` 模式定点修复，不等 LLM 审计。
+
+### 审计-修订闭环加固
+
+实测发现 `rewrite` 模式引入 6 倍 AI 标记词，现在：
+
+- 自动修订模式从 `rewrite` 改为 `spot-fix`（只改问题句，不碰其余正文）
+- 修订后对比 AI 标记数，如果修订反而增多 AI 痕迹，丢弃修订保留原文
+- 再审温度锁 0（消除审计随机性，同一章不再出现 0-6 个 critical 的波动）
+- `polish` 模式加固边界（禁止增删段落、改人名、加新情节）
+
+### 其他 v0.4 变更
+
+- 审计维度从 26 扩展到 32（+4 番外维度 + dim 27 敏感词 + dim 32 读者期待管理）
+- 审计员联网搜索：年代考据题材可调用 DuckDuckGo 核实真实事件/人物/地理
+- 调度器重写：AI 节奏（默认 15 分钟一轮）、并行书处理、立即重试、每日上限
+- 修订者新增 `spot-fix` 模式（定点修复）
+- `book_rules.md` 的 `additionalAuditDimensions` 支持中文名称匹配
+- 全部 5 个题材激活 dim 24-26（支线停滞/弧线平坦/节奏单调）
+
+---
+
 ## v0.3 更新
 
 创作规则三层分离 + 跨章记忆 + AIGC 检测 + Webhook。
@@ -112,13 +191,13 @@ fatigueWordsOverride: ["瞳孔骤缩", "不可置信"]   # 覆盖题材默认
 
 主角人设锁定、数值上限、自定义禁令、疲劳词覆盖——每本书的规则独立调整，不影响题材模板。
 
-### 26 维度审计
+### 32 维度审计
 
-审计细化为 26 个维度，按题材自动启用对应的子集：
+审计细化为 32 个维度，按题材自动启用对应的子集：
 
-OOC检查、时间线、设定冲突、伏笔、节奏、文风、信息越界、词汇疲劳、利益链断裂、配角降智、配角工具人化、爽点虚化、台词失真、流水账、知识库污染、视角一致性、战力崩坏、数值检查、年代考据、段落等长、套话密度、公式化转折、列表式结构、支线停滞、弧线平坦、节奏单调
+OOC检查、时间线、设定冲突、战力崩坏、数值检查、伏笔、节奏、文风、信息越界、词汇疲劳、利益链断裂、年代考据、配角降智、配角工具人化、爽点虚化、台词失真、流水账、知识库污染、视角一致性、段落等长、套话密度、公式化转折、列表式结构、支线停滞、弧线平坦、节奏单调、敏感词检查、正传事件冲突、未来信息泄露、世界规则跨书一致性、番外伏笔隔离、读者期待管理
 
-玄幻/仙侠全 26 维度，都市 24 维度（含年代考据），恐怖 22 维度。其中 dim 20-23（AI 痕迹）由纯规则引擎检测，不消耗 LLM 调用。
+dim 20-23（AI 痕迹）+ dim 27（敏感词）由纯规则引擎检测，不消耗 LLM 调用。dim 28-31（番外维度）检测到 `parent_canon.md` 自动激活。dim 32（读者期待管理）始终开启。
 
 ### 去 AI 味
 
@@ -135,7 +214,7 @@ OOC检查、时间线、设定冲突、伏笔、节奏、文风、信息越界
 ### 其他
 
 - 支持 OpenAI + Anthropic 原生 + 所有 OpenAI 兼容接口
-- 修订者支持 polish / rewrite / rework / anti-detect 四种模式
+- 修订者支持 polish / rewrite / rework / anti-detect / spot-fix 五种模式
 - 无数值系统的题材不生成资源账本
 - 所有命令支持 `--json` 结构化输出，OpenClaw / 外部 Agent 可直接解析
 - book-id 自动检测：项目只有一本书时省略 book-id
@@ -317,6 +396,7 @@ inkos up                       # 守护进程模式
 | `inkos detect [id] [n]` | AIGC 检测（`--all` 全部章节，`--stats` 统计） |
 | `inkos style analyze <file>` | 分析参考文本提取文风指纹 |
 | `inkos style import <file> [id]` | 导入文风指纹到指定书 |
+| `inkos import canon [id] --from <parent>` | 导入正传正典到番外书 |
 | `inkos update` | 更新到最新版本 |
 | `inkos up / down` | 启动/停止守护进程 |
 
@@ -376,9 +456,9 @@ inkos up                       # 守护进程模式
 inkos/
 ├── packages/
 │   ├── core/              # Agent 运行时、管线、状态管理
-│   │   ├── agents/        # architect, writer, continuity, reviser, radar, ai-tells, detector, style-analyzer
+│   │   ├── agents/        # architect, writer, continuity, reviser, radar, ai-tells, post-write-validator, sensitive-words, detector, style-analyzer
 │   │   ├── pipeline/      # runner, agent (tool-use), scheduler, detection-runner
-│   │   ├── state/         # 基于文件的状态管理器（7 真相文件 + 快照）
+│   │   ├── state/         # 基于文件的状态管理器（7+ 真相文件 + 快照）
 │   │   ├── llm/           # OpenAI + Anthropic 双 SDK 接口 (流式)
 │   │   ├── notify/        # Telegram, 飞书, 企业微信, Webhook
 │   │   └── models/        # Zod schema 校验
@@ -403,12 +483,16 @@ TypeScript 单仓库，pnpm workspaces 管理。
 - [x] 可插拔雷达（RadarSource 接口）
 - [x] 外部 Agent 集成（OpenClaw 等）
 - [x] 题材自定义 + 单本书规则（genre CLI + book_rules.md）
-- [x] 26 维度连续性审计（含 AI 痕迹检测）
+- [x] 32 维度连续性审计（含 AI 痕迹检测 + 番外维度）
 - [x] 去 AI 味铁律 + 文风指纹注入
 - [x] 多 LLM provider（OpenAI + Anthropic + 兼容接口）
 - [x] AIGC 检测 + 反检测改写管线
 - [x] Webhook 通知 + 智能调度（质量门控）
 - [x] 跨章节连贯性（章节摘要 + 支线/情感/角色矩阵）
+- [x] 番外写作（正典导入 + 4 维度审计 + 信息边界管控）
+- [x] 文风仿写（统计指纹 + LLM 风格指南 + 写手注入）
+- [x] 写后验证器（11 条硬规则 + 自动 spot-fix）
+- [x] 审计-修订闭环加固（AI 标记守卫 + 温度锁）
 - [ ] `packages/studio` Web UI 审阅编辑界面
 - [ ] 多模型路由（不同 agent 用不同模型）
 - [ ] 自定义 agent 插件系统

package.json

@@ -1,6 +1,6 @@
 {
   "name": "inkos",
-  "version": "0.3.6",
+  "version": "0.4.0",
   "private": true,
   "description": "Multi-agent novel production system",
   "license": "MIT",

packages/cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@actalk/inkos",
-  "version": "0.3.6",
+  "version": "0.4.0",
   "description": "InkOS CLI — multi-agent novel production system",
   "type": "module",
   "bin": {

packages/core/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@actalk/inkos-core",
-  "version": "0.3.6",
+  "version": "0.4.0",
   "description": "InkOS core: models, agents, pipeline, LLM integration",
   "type": "module",
   "main": "dist/index.js",