Runtime Environment Compatibility: Claude Code / OpenCode / Codex / ZCode / OpenClaw are built-in adaptation targets; environments that can read project files such as NarraFork, Web AI, and custom Agents can execute the long-form workflow according to this skill. When checking professional agents, search in the order of

.claude/agents/{agent}.md

→

.opencode/agents/{agent}.md

→

.codex/agents/{agent}.toml

; if not found, Codex returns

unknown agent_type

, or

.zcode/

is detected (ZCode 3.3.4 does not execute project custom agents), execute directly in solo/direct mode and report fallback.

Spawn Version Notice (does not block spawn): First read

agents_version

from the project root

.story-deployed

. When it is inconsistent with this version's

agents_version: 25

(missing marker, missing/non-integer field, less than or greater than 25) proceed with file existence check and spawn as usual, while reporting

Notice: agents bundle version mismatch (project {N}, current version 25)

and prompting to re-run

/story-setup

before starting a new session; if greater than 25, additionally prompt to update oh-story-claudecode first and do not use local old setup to downgrade and overwrite. Only fall back to solo/direct mode when agent files are missing or the runtime does not expose custom agents, and report

Fallback: ... -> solo

.

1. Define emotion first, then the story. Every scene must serve a clear emotional goal. Scenes that cannot explain what emotion they deliver should not exist.
2. Start from proven patterns. First ask "what has been proven effective, how can I re-deliver it", rather than directly starting from "what I want to write". Scan rankings for directions, analyze existing novels for modules, and reference benchmarks for pacing.
3. Assemble with modules, don't reinvent the wheel. Every genre has proven plot patterns—how to lay out reversals, how to deliver satisfying moments, how to build emotional tension. Find the right modules, treat specific characters from benchmark novels as functional positions (rival/ally/catalyst), then map them to your own characters. Fill these functional positions with your own material.
4. Load only necessary information. When writing each chapter, only load information that "you would write incorrectly without knowing it". This includes character statuses, pending foreshadowing, and relevant settings. Keep the rest in the file system.
5. Contract and progression decisions follow authoritative reference files. When involving reader contracts, protagonist agency, interest security, expectation debts, endgame reserves (endgame trump cards/upgrade stages), organization/force boundaries, and risk assessments of contract security/needs reinforcement/contract breach, first calibrate with

   references/reader-contract-and-progression.md

   , do not copy long rules within SKILL.md.

Reverse-match genre from emotion: If the user first mentions an emotional feeling but not a genre, reverse-match from the above table—for example, "Release of Satisfaction" points to Face-slapping/Comeback, then find sub-directions under this genre in

genre-catalog.md

.

references/workflow-setup.md

references/workflow-daily.md

references/workflow-revision.md

Start New Volume: If a new volume introduces new characters/forces/settings, first return to Phase 2 for incremental supplementation, then proceed to Phase 3 to supplement the new volume's detailed outline, finally enter Phase 4 for writing. If it is a pure continuation, directly return to Phase 3.

/story-long-write

$story-long-write

• Empty project → Suggest saying "Help me start a novel" or provide

  选题决策.md

  first;
• Has settings/outline but no full text → Suggest saying "Write Chapter 1", "Only write 1 chapter" or "Daily update 2 chapters";
• Has full text + tracking → Display the last completed chapter and the status of the next chapter's detailed outline, suggest saying "Daily update 3 chapters", "Only write 1 chapter", "Confirm chapter by chapter" or "Revise Chapter X".

references/workflow-daily.md

Analysis Library/Benchmark Relationship:

拆文库/

= Original output of analyze skill, serves as data source.

对标/

= Reference view of writing project, stores subset of benchmark data related to this project. When referencing a benchmark novel for the first time, copy relevant subdirectories (chapters/characters/plots/settings),

剧情/节奏.md

,

剧情/情绪模块.md

,

文风.md

and

拆文报告.md

from

拆文库/{书名}/

to

对标/{书名}/

.

Benchmark Novel Path Search: Prioritize

{项目}/对标/{书名}/

, if not found fall back to

拆文库/{书名}/

. All benchmark data loading below follows this rule.

{书名}/
├── 设定/
│   ├── 世界观/
│   │   ├── 背景设定.md        # 时代背景、地理、历史
│   │   ├── 力量体系.md        # 修炼/能力/等级体系
│   │   └── ...
│   ├── 角色/
│   │   ├── 沈栀.md            # 每个人物一个文件，文件名用角色名
│   │   └── ...
│   ├── 势力/
│   │   ├── 天机阁.md          # 每个势力/组织一个文件
│   │   └── ...
│   ├── 关系.md                # 角色关系映射
│   ├── 题材定位.md            # 题材核心梗+对标分析+终局底牌/升级台阶（防写无可写）
│   └── 题材正文提示卡.md       # 题材正文核心：边界/期待/爽点/节奏/禁漂移
├── 大纲/
│   ├── 大纲.md                # 全书卷级结构
│   ├── 卷纲_第一卷.md         # 每卷一个：对标结构坐标+剧情单元+情绪弧线(含章节定位)+人物弧线+伏笔+反转
│   └── 细纲_第001章.md        # 每章一个：章节定位+事件+钩子(按章节定位,章首/章尾/段落级)+爽点+悬念
├── 正文/
│   ├── 第001章_章名.md
│   └── ...
├── 对标/                          ← 拆文产出的结构化资产
│   └── {对标书名}/
│       ├── 原文/
│       │   ├── 第001章_章名.md
│       │   └── ...
│       ├── 角色/                  ← 从拆文库/结构化输出同步
│       │   └── {角色名}.md
│       ├── 剧情/                  ← 从拆文库/结构化输出同步
│       │   ├── {剧情单元名}.md
│       │   ├── 故事线.md
│       │   ├── 节奏.md             # 关键信息推进 + 情绪触动点 + 爆发节奏（权威节奏索引）
│       │   └── 情绪模块.md         # 读者需求/情绪引擎 + 可复现模块（权威模块索引）
│       ├── 设定/                  ← 从拆文库/结构化输出同步
│       │   ├── 世界观/             ← 按主题拆分到子目录
│       │   │   ├── 背景设定.md
│       │   │   ├── 力量体系.md
│       │   │   ├── 地理.md
│       │   │   └── 金手指.md
│       │   └── 势力/
│       │       └── {势力名}.md
│       └── 拆文报告.md
├── 追踪/
│   ├── _tracking-state.json        ← 唯一结构化权威状态
│   ├── 上下文.md                  ← 派生续写状态卡（固定 7 栏），≤12KB
│   ├── 逐章记录/第NNN章.md          ← 未来相关紧凑记录，≤3072 字节
│   ├── 角色状态/{角色名}.md         ← 派生核心角色当前快照
│   ├── 伏笔.md                    ← 派生伏笔当前视图
│   └── 时间线/{作者真相.md,读者已知.md}
├── 参考资料/
│   └── {topic}.md             # story-researcher 输出的研究资料

primary benchmark novel

genre-prose-cards.md

genre-prose-cards/

style-genre-modules.md

tracking_commit.py check

设定/角色/

1. Missing character status file → For current protocol projects, first run

   tracking_commit.py check

   , then re-run the complete transaction that generated this status; if

   _tracking-state.json

   is missing but full text exists, re-run

   /story-import

   . Do not directly hand-write snapshots by temporary inference from previous content.
2. Missing subdirectories for non-main artifacts such as characters, ordinary plot units or settings → Search project view and root directory data source according to "Benchmark Novel Path Search", skip this optional module if still missing. This rule does not apply to

   剧情/情绪模块.md

   and

   剧情/节奏.md

   .
3. Missing

   剧情/情绪模块.md

   /

   剧情/节奏.md

   → Must stop pre-writing preparation, set

   missing_primary_contract: true

   and provide

   repair_action

   : re-run

   /story-long-analyze

   Stage 3+ or re-run

   /story-import

   , do not use summary files to pretend authoritative modules have been recalled.
4. Missing

   文风.md

   but has benchmark novel → If

   设定/文风.md

   (with substantive content) exists, continue with custom writing style mode; otherwise, daily update writing style recall fails fast, prompt to first run

   /story-long-analyze

   Stage 6 and

   /story-import

   to synchronize. Projects with no benchmarks skip writing style recall and do not block (use

   设定/文风.md

   for writing if it exists). Emotion/rhythm axis (

   missing_primary_contract

   ) is independent, custom writing style mode does not exempt its fail-fast.
5. Missing foreshadowing/timeline files → Consider current semantic checkpoint damaged, stop writing full text; first run

   tracking_commit.py check

   , then repair with transactions. Plans in volume outline/outline cannot replace current checkpoint of occurred facts.
6. Missing

   设定/题材正文提示卡.md

   → Do not block; before writing, accurately match index in

   references/genre-prose-cards.md

   from

   设定/题材定位.md

   , and only read corresponding single genre card in

   references/genre-prose-cards/

   (mark high/medium/low confidence as original card), if no match, generate short

   genre_prose_card

   immediately with general genre modules in

   references/style-genre-modules.md

   . Only when

   设定/题材定位.md

   is also missing, fall back to detailed outline and target platform to create low-confidence genre card, and state clearly in intent confirmation.

1. 剧情/情绪模块.md

   is the authoritative source for reader needs / emotion engine, satisfying novel routine frameworks, reproducible modules and recombination guides.

2. 剧情/节奏.md

   is the authoritative source for key information advancement, chapter expansion technique aggregation, emotional touch points and outbreak rhythm.

3. 文风.md

   only manages sentence length, punctuation, subtext in dialogue, original text anchors and other styles; it cannot override emotion modules or rhythm intent. Custom writing style

   设定/文风.md

   (written by user, not overwritten by import/analysis) has higher priority than benchmark

   文风.md

   : when it contains substantive content, it serves as authoritative style base, benchmark writing style is demoted to reference and sentence length value fallback; writing methods that hit hard safety lines (

   ……

   / em dash / blank line between paragraphs / fragmented sentences) are still normalized according to narrative-writer, custom only takes over sentence length / soft punctuation / subtext / emotion alternation.

4. 章节/第K章_摘要.md

   is specific chapter evidence, used to verify and supplement authoritative indexes, does not reverse-override

   情绪模块.md

   /

   节奏.md

   .

5. 拆文报告.md

   ,

   剧情/故事线.md

   are projections/summaries; if conflicting with

   剧情/情绪模块.md

   or

   剧情/节奏.md

   , writing follows the two authoritative files, and record conflict sources in pre-writing preparation

   gaps.conflict

   .

• One file per character:

  角色/角色名.md

  , convenient for on-demand reading
• One file per force:

  势力/势力名.md

  , for organizations/sects/families/countries, etc.
• Worldbuilding split by topic: Background, power system, social structure, etc., are independent
• One detailed outline file per chapter:

  细纲_第XXX章.md

  , includes hook design, one-to-one correspondence with full text
• Full text split by chapter: One file per chapter,

  第XXX章_章名.md

• Write directly to

  正文/

  directory after completing each chapter, do not output to conversation first

• Respond in the user's language; use the same language as the user's input
• Chinese responses follow the "Chinese Copywriting Typesetting Guide"