Skip to content
Back to search

qiaomu-novel-generator

Generate original, complete, highly gripping Chinese short fiction from a story theme, character setup, synopsis, trope, classic-plot inspiration, or draft excerpt. Use when the user asks for 小说, 短篇小说, 故事, 网文样章, 开篇钩子, 悬念改写, 人物欲望强化, 情节升级, 续写, 经典桥段重构, or wants a blend of excellent writers' general narrative techniques without copying protected text or imitating a living author's distinctive style.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/joeseesun--qiaomu-novel-generator/badge.svg)](/listings/joeseesun--qiaomu-novel-generator)

HTML

<a href="/listings/joeseesun--qiaomu-novel-generator"><img src="/api/listings/joeseesun--qiaomu-novel-generator/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

一行安装:

npx skills add joeseesun/qiaomu-novel-generator

确认能被发现:

npx skills add joeseesun/qiaomu-novel-generator --list

确认本地已安装:

test -f .agents/skills/qiaomu-novel-generator/SKILL.md || test -f ~/.agents/skills/qiaomu-novel-generator/SKILL.md

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 145 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. ---
  2. name: qiaomu-novel-generator
  3. description: |
  4. Generate original, complete, highly gripping Chinese short fiction from a story theme, character setup, synopsis, trope, classic-plot inspiration, or draft excerpt. Use when the user asks for 小说, 短篇小说, 故事, 网文样章, 开篇钩子, 悬念改写, 人物欲望强化, 情节升级, 续写, 经典桥段重构, or wants a blend of excellent writers' general narrative techniques without copying protected text or imitating a living author's distinctive style.
  5. ---
  6. # Qiaomu Novel Generator
  7. 把一句主题、人物设定、梗概、经典作品信号或已有片段,先通过必要的搜索/桥段拆解收敛成吸引人的剧情策略和大纲,再生成原创、完整、强钩子、高张力、低 AI 味的中文短篇小说。
  8. Copyright (c) 向阳乔木
  9. X: https://x.com/vista8
  10. GitHub: https://github.com/joeseesun/
  11. ## Operating Mode
  12. Run as a production-lite fiction writing skill.
  13. Default assumptions:
  14. - The user ultimately wants a complete Chinese short story, but a vague premise should be clarified through compact choices before drafting.
  15. - Default output length is 1800-4000 Chinese characters unless the user asks for a different length.
  16. - Default goal is reader compulsion: make the first page impossible to ignore, then keep raising the cost of every choice.
  17. - Treat named writers, films, TV shows, novels, and famous scenes as craft signals, not copy targets. Convert them into general techniques such as suspense, restraint, moral pressure, voice economy, scene rhythm, reversal, public proof, and emotional payoffs.
  18. - Do not directly imitate a living author's distinctive style. Do not copy protected passages, famous scenes, character names, signature lines, setting names, or recognizable plot sequences.
  19. - When the user gives a plot idea, proactively offer several classic beat inspirations and let the user choose or combine before outlining.
  20. - When the user asks to search first or names a specific modern work whose public context matters, use search if available, extract only public premise/craft signals, cite sources when reporting, and then remix through abstract story functions.
  21. - If the user provides an existing draft, preserve useful facts, promises, and character intent; rewrite the narrative engine instead of merely polishing adjectives.
  22. - Avoid formulaic AI wording in visible prose, especially `不是X,而是Y`, unless it is a deliberate character voice.
  23. - A strong opening hook must not create the wrong genre contract. If a line sounds supernatural, impossible, or metaphorical, immediately anchor the literal situation unless the story is intentionally supernatural.
  24. ## Hooked Workflow
  25. Use hooks as fixed checkpoints. They are conceptual hooks, not mandatory runtime APIs. Their job is to keep the skill evolvable without hard-coding one user's feedback or one story's content into the core rules.
  26. 1. Intent Hook: identify the input type, target genre, reader promise, premise clarity, and whether the user has explicitly approved drafting.
  27. 2. Source Research Hook: when the user explicitly asks to search or names a reference that needs current/public context, use `references/source-research-remix.md` to gather 3-6 public signals and reduce them to a Research Intake Card.
  28. 3. Inspiration Remix Hook: when the user gives a plot, genre, trope, or named work/writer, use `references/inspiration-remix-playbook.md` to offer classic beat/style-signal choices before outlining. Reduce references to functions and craft sliders.
  29. 4. Story Engine Library Hook: select the primary emotional payoff, high-pressure relationship, conflict arena, 2-4 plot engines, escalation ladder, and hook mode from `references/story-engine-library.md`.
  30. 5. Prewrite Interview Hook: if the request is vague, follow `references/prewrite-interview.md` and ask with short numbered choices. Do not draft the full story yet.
  31. 6. Story Strategy Hook: if the premise is usable but the outline is not confirmed, provide classic inspiration options, `小说如何吸引人`, and a compact outline, then wait for confirmation unless the user explicitly asked to skip discussion.
  32. 7. Story Engine Hook: after confirmation, extract:
  33. - protagonist desire
  34. - visible obstacle
  35. - hidden pressure
  36. - moral or emotional cost
  37. - reader promise
  38. - final aftertaste
  39. 8. Technique Hook: choose 3-5 technique engines from `references/technique-matrix.md`. Use a mix, not a stack of author imitations.
  40. 9. Plan Hook: build a compact story plan:
  41. - first disturbance within the first 3 paragraphs
  42. - protagonist makes an active choice
  43. - conflict escalates at least 3 times
  44. - each scene reveals one new fact or removes one safe option
  45. - final turn reframes the opening
  46. 10. Draft Hook: draft the complete short story according to `references/output-contract.md`.
  47. 11. Anti-AI Language Hook: apply `references/anti-ai-language.md`; rewrite formulaic narration before returning.
  48. 12. Quality Hook: self-check against `references/quality-checklist.md` and the relevant rubric in `references/genre-quality-rubric.md`. Revise before returning if the story fails on hook, desire, escalation, dialogue, image, reversal/suspense, ending aftertaste, opening clarity, or anti-AI language.
  49. 13. Feedback Hook: when the user gives critique, classify the failure mode before rewriting. Do not only patch the current paragraph.
  50. 14. Evolution Hook: only promote feedback into stable skill rules when it is repeated, high-signal, or fixes a transferable failure mode. See `references/evolution-loop.md`.
  51. 15. If writing a sample, deliverable, or file artifact, run `python3 scripts/validate_skill.py` from the skill directory before calling it done. When a draft file exists, optionally run `python3 scripts/evaluate_story.py <draft-file>` for a lightweight metrics report.
  52. ## Output Defaults
  53. For a vague new-story request, output a compact option card first. The user should be able to reply `按默认` or `1B 2A 3C`.
  54. For a usable premise that has not been confirmed, output:
  55. ```text
  56. ## 经典桥段启发
  57. [3-6 options reduced to reusable functions]
  58. ## 小说如何吸引人
  59. [strategy bullets]
  60. ## 大纲
  61. [compact outline]
  62. 确认后我再写正文。
  63. ```
  64. After the user confirms, or if the user explicitly asks to skip discussion and write directly, output:
  65. ```text
  66. 《标题》
  67. [完整小说正文]
  68. ```
  69. Only add a visible `创作自检` section when the user asks for evaluation, asks for a reusable workflow, or the output is a sample/benchmark artifact.
  70. If the user asks for "按某作家风格", briefly translate the request into craft terms before writing:
  71. ```text
  72. 我会使用可泛化技法:短句留白、对白推进、危险感、命运压力和反转结构;不复刻具体作者的独特表达或已有情节。
  73. ```
  74. Then write the story.
  75. If the user names a living author, do not promise exact imitation. Translate to craft features and continue with an original story.
  76. ## Revision Modes
  77. Use the same skill for:
  78. - `开篇更抓人`: replace summary opening with immediate disturbance, concrete image, and unanswered danger.
  79. - `人物动机更强`: give the protagonist a visible want and a private wound; make both collide.
  80. - `冲突升级`: add public pressure, time limit, betrayal, cost, or impossible choice.
  81. - `经典桥段重构`: offer famous film/TV/novel beat cards, let the user choose, then remix into new characters, setting, stakes, objects, and ending.
  82. - `风格技法转译`: convert named-author or named-work requests into craft sliders; avoid direct living-author style imitation.
  83. - `对白更有张力`: remove explanation; make each line hide intent, threat, desire, or reversal.
  84. - `结尾更有余味`: make the last image repay an earlier detail while opening a larger question.
  85. - `去 AI 味`: remove formulaic contrast, teaching-tone transitions, decorative dashes, and theme explanations; replace them with action, image, dialogue, and consequence.
  86. - `开篇误读`: if a poetic or compressed hook makes readers misunderstand the literal event or genre, rewrite with a concrete anchor in the same sentence or next sentence.
  87. - `改成完整短篇`: convert outline or fragment into a beginning-middle-end story, not a synopsis.
  88. - `用户反馈迭代`: classify the critique, choose one dominant failure mode, rewrite against that failure, then decide whether the lesson is only task-local or should be added to skill references.
  89. ## Boundaries
  90. - Do not copy protected text, famous scenes, unique names, signature artifacts, or recognizable plot beats from copyrighted fiction.
  91. - Do not copy a recognizable sequence of beats from one source. If using inspiration, change at least setting, relationship, stakes, object/rule, and ending.
  92. - Do not present a story as "in the exact style of" a living author. Transform the request into general craft features.
  93. - Do not use real private persons as fictional criminals, villains, abusers, or scandal subjects without clear fictionalization and safety framing.
  94. - Do not output sexual content involving minors, explicit sexual coercion, instructions for real violence, or content that glamorizes criminal abuse.
  95. - For sensitive requests, pivot to fictionalized, non-instructional, emotionally focused storytelling.
  96. ## Reference Files
  97. - `references/technique-matrix.md`: technique engines and author-signal-to-craft translation.
  98. - `references/source-research-remix.md`: search-first protocol for turning public source context into abstract remix cards.
  99. - `references/inspiration-remix-playbook.md`: classic film/TV/novel beat cards, style-signal translation, and remix boundary.
  100. - `references/story-engine-library.md`: emotional payoff types, high-pressure relationships, plot engines, escalation ladders, arenas, and hook bank.
  101. - `references/prewrite-interview.md`: option-based story clarification, strategy sheet, and outline confirmation rules.
  102. - `references/output-contract.md`: story output shape, length, and drafting rules.
  103. - `references/anti-ai-language.md`: anti-trope language gate for formulaic AI wording.
  104. - `references/quality-checklist.md`: seven-part self-check and repair rules.
  105. - `references/genre-quality-rubric.md`: genre-aware quality rubrics for loading the right reader promise.
  106. - `references/evolution-loop.md`: hook-based feedback, rule promotion, and anti-overfitting process.
  107. - `examples/sample-01-wuxia-suspense.md`:江湖悬疑完整短篇样例与自评。
  108. - `examples/sample-02-sci-fi-memory.md`:近未来科幻完整短篇样例与自评。
  109. - `scripts/validate_skill.py`: local structural validator for required files and sample self-evaluations.
  110. - `scripts/evaluate_story.py`: optional draft metrics reporter for abstraction, dialogue, scene, and hook signals.
README.md · README · 370 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. # qiaomu-novel-generator
  2. > 你想写一个爽文、武侠、修仙、悬疑或现代组织内斗故事,AI 却常常给你一篇“像大纲、像说明书、像套路拼贴”的东西。
  3. > `qiaomu-novel-generator` 先帮你把剧情钩子、经典桥段、人物欲望、冲突升级和结尾余味定住,再写成原创、完整、低 AI 味的中文小说。
  4. >
  5. > Turn a rough Chinese story idea into a selectable plot strategy, then into a complete gripping short story.
  6. [![Install](https://img.shields.io/badge/install-npx%20skills%20add-0f766e?style=for-the-badge)](#安装)
  7. [![Stars](https://img.shields.io/github/stars/joeseesun/qiaomu-novel-generator?style=for-the-badge&logo=github)](https://github.com/joeseesun/qiaomu-novel-generator/stargazers)
  8. [![Forks](https://img.shields.io/github/forks/joeseesun/qiaomu-novel-generator?style=for-the-badge&logo=github)](https://github.com/joeseesun/qiaomu-novel-generator/forks)
  9. [![Issues](https://img.shields.io/github/issues/joeseesun/qiaomu-novel-generator?style=for-the-badge&logo=github)](https://github.com/joeseesun/qiaomu-novel-generator/issues)
  10. [![Last commit](https://img.shields.io/github/last-commit/joeseesun/qiaomu-novel-generator?style=for-the-badge&logo=git)](https://github.com/joeseesun/qiaomu-novel-generator/commits/main)
  11. [![License](https://img.shields.io/github/license/joeseesun/qiaomu-novel-generator?style=for-the-badge)](LICENSE)
  12. ```bash
  13. npx skills add joeseesun/qiaomu-novel-generator
  14. ```
  15. **中文** | [English](#english)
  16. ## 为什么值得装
  17. 普通小说生成器经常直接开写。结果是:
  18. - 开头没有钩子,读者翻两段就走。
  19. - 主角没有欲望,只是在被剧情推着跑。
  20. - 爽点来得太便宜,打脸不够爽。
  21. - 想“参考某部作品”,最后变成危险的照搬。
  22. - 想“学某位作者”,最后只学到表面腔调。
  23. - 越改越像 AI:`不是X,而是Y`、总结腔、教学腔、黑话堆满。
  24. 这个 skill 的思路相反:**先讨论剧情怎么吸引人,再写正文。**
  25. 它会先把你的输入拆成可选择的剧情引擎:
  26. ```text
  27. 情绪承诺 -> 高压关系 -> 冲突场 -> 经典桥段重构 -> 升级节奏 -> 结尾余味
  28. ```
  29. 你可以回复 `按默认`、`更爽一点`、`更悬疑`、`1B 2A 3C`,它再进入正文写作。
  30. ## 真实输出预览
  31. | 输入 | 输出 |
  32. |---|---|
  33. | 江湖悬疑、命运压力、交易与误判 | [《雨夜验剑》](examples/sample-01-wuxia-suspense.md) |
  34. | 近未来科幻、记忆交易、身份反转 | [《第七枚记忆》](examples/sample-02-sci-fi-memory.md) |
  35. 你还可以让它先给剧情策略,而不是直接写:
  36. ```text
  37. 我想写一个类似《凡人修仙传》的小说。
  38. ```
  39. 它会先给你:
  40. - 底层小人物的生存爽点
  41. - 地下交易、杀人夺宝、宗门小比等可选桥段
  42. - 主角欲望、隐藏压力和升级代价
  43. - 可确认的大纲
  44. - 确认后再写第一章或完整短篇
  45. ## 它最擅长什么
  46. ### 1. 把一句想法变成可追读剧情
  47. 你说:
  48. ```text
  49. 帮我生成一个小说。
  50. ```
  51. 它不会马上塞给你一篇泛泛的故事,而是先给紧凑选项:
  52. ```text
  53. 1. 主情绪:A 逆袭爽 / B 悬疑惊 / C 黑色反转
  54. 2. 高压关系:A 外来者/上位者 / B 仇人/合作者 / C 救命恩人/错认者
  55. 3. 冲突场:A 公开审查 / B 地下拍卖 / C 家宴质证
  56. ```
  57. ### 2. 把经典桥段拆成“功能”,再重构
  58. 你说:
  59. ```text
  60. 用《食神》那种跌落神坛再公开翻身的结构,换成修仙故事。
  61. ```
  62. 它会借“失势、底层重学、公开回归”的结构,不复制人物、台词、场面和情节链。
  63. 可用的桥段卡包括:
  64. - 跌落神坛再翻身
  65. - 地下拍卖/鉴宝反转
  66. - 法庭/公开质证翻案
  67. - 隐藏身份与迟来承认
  68. - 后台项目视角与组织讽刺
  69. - 宗门试炼、规则漏洞、公开打脸
  70. ### 3. 搜索后重构,而不是复述原作
  71. 你说:
  72. ```text
  73. 先联网搜一下这本书,再改成大厂内斗爽文。
  74. ```
  75. 它会先提取公开资料里的稳定事实、读者爽点、叙事功能和不可复制边界,再给你桥段组合。
  76. 它不把原作角色、名场面、签名句、世界观设定或连续情节链搬进新故事。
  77. ### 4. 写之前先定大纲
  78. 适合你想认真写一个故事的时候:
  79. ```text
  80. 我想写一个类似冰与火之歌的现代职场剧情,我们先讨论下。
  81. ```
  82. 它会先输出:
  83. - 经典桥段启发
  84. - 小说如何吸引人
  85. - 主角欲望
  86. - 隐藏压力
  87. - 冲突升级
  88. - 章节/场景钩子
  89. - 结尾余味
  90. - 大纲
  91. 确认后再写正文。
  92. ### 5. 专门处理“AI 味”
  93. 内置反 AI 味检查:
  94. - 避免 `不是X,而是Y`
  95. - 避免 `关键在于`、`值得注意的是`
  96. - 避免总结式结尾
  97. - 避免装饰性破折号
  98. - 避免开篇为了诗性而误导题材
  99. 例如它会把:
  100. ```text
  101. 死人走进了酒肆。
  102. ```
  103. 改成更清楚的危险画面:
  104. ```text
  105. 一个快死的人走进了酒肆,胸口的血已经在狐裘上结成冰。
  106. ```
  107. ## 安装
  108. 一行安装:
  109. ```bash
  110. npx skills add joeseesun/qiaomu-novel-generator
  111. ```
  112. 确认能被发现:
  113. ```bash
  114. npx skills add joeseesun/qiaomu-novel-generator --list
  115. ```
  116. 确认本地已安装:
  117. ```bash
  118. test -f .agents/skills/qiaomu-novel-generator/SKILL.md || test -f ~/.agents/skills/qiaomu-novel-generator/SKILL.md
  119. ```
  120. ## 前置条件
  121. - [ ] 已安装 Node.js:运行 `node -v` 检查;没有的话可从 https://nodejs.org 安装。
  122. - [ ] 已安装 npm/npx:运行 `npx --version` 检查;Node.js 通常会自带。
  123. - [ ] 使用支持 agent skills 的工具,例如 Codex、Claude Code 或其他兼容本地 skills 的 Agent。
  124. - [ ] 如果要让它联网研究参考作品,当前 Agent 环境需要允许浏览或搜索。
  125. ## 你可以这样说
  126. ```text
  127. 用 qiaomu-novel-generator,写一个不会武功的书生误入江湖死局的完整短篇。
  128. ```
  129. ```text
  130. 帮我生成一个小说,先给几个方向让我选。
  131. ```
  132. ```text
  133. 我想写一个类似《凡人修仙传》的小说,先给爽点和大纲。
  134. ```
  135. ```text
  136. 给我几个著名电影、电视剧、小说的经典桥段选择,组合成一个新故事。
  137. ```
  138. ```text
  139. 先联网搜一下《太白金星有点烦》的公开资料,再改成现代组织内斗爽文。
  140. ```
  141. ```text
  142. 这个开篇有点误导,我以为是鬼怪,保留钩子但写清楚发生了什么。
  143. ```
  144. ```text
  145. 这段 AI 味太重,尤其是“不是……而是……”,帮我改得像人写的。
  146. ```
  147. ## 工作流
  148. ```mermaid
  149. flowchart TD
  150. A["用户给主题 / 作品信号 / 片段"] --> B["识别读者承诺"]
  151. B --> C["经典桥段启发"]
  152. C --> D["剧情引擎组合"]
  153. D --> E["小说如何吸引人"]
  154. E --> F["大纲确认"]
  155. F --> G["完整正文"]
  156. G --> H["去 AI 味 / 开篇清晰 / 质量自检"]
  157. ```
  158. ## 内置剧情引擎
  159. | 类型 | 适合制造什么爽感 |
  160. |---|---|
  161. | 隐藏身份 | 被低估后公开翻盘 |
  162. | 地下拍卖/鉴宝 | 废物变宝、价值误判、反向打脸 |
  163. | 宗门试炼/公开审查 | 规则漏洞、见证人反应、资历压制反杀 |
  164. | 杀人夺宝/禁忌交易 | 机缘带来危险,危险逼出成长 |
  165. | 双强博弈 | 互相试探、互相利用、反败为胜 |
  166. | 组织内斗 | 抢功、甩锅、文件证据、公开复盘 |
  167. | 悬疑线索链 | 每个答案打开更坏的问题 |
  168. | 追妻/错认/迟来承认 | 情绪拉扯、尊严、代价选择 |
  169. ## 质量门槛
  170. 每篇故事都会按这些维度自检:
  171. - 开篇钩子:前三段有危险、羞辱、损失或谜题。
  172. - 人物欲望:主角马上想要一个看得见的东西。
  173. - 冲突升级:至少三次变难,不靠流水账。
  174. - 对白张力:对白里有威胁、试探、隐瞒或交换。
  175. - 画面感:用物件、动作、场景承载信息。
  176. - 反转/悬念:前文细节在后文换意义。
  177. - 结尾余味:解决当下问题,同时留下回响。
  178. - 去 AI 味:删掉公式化对比和总结腔。
  179. - 开篇语义清晰:钩子不能让读者误判题材或基本事实。
  180. ## 安全边界
  181. 这个 skill 支持“借鉴结构”,不支持“复制作品”。
  182. 允许:
  183. - 借悬念、留白、命运压力、类型爽点、结构反转。
  184. - 把经典桥段拆成通用功能再组合。
  185. - 把作者/作品信号翻译成可泛化技法。
  186. 不做:
  187. - 不复制受版权保护文本。
  188. - 不搬运名场面、角色、地名、招式、签名句。
  189. - 不复刻可识别的连续情节链。
  190. - 不直接仿写在世作者的独特个人文风。
  191. ## 开发与验证
  192. ```bash
  193. cd ~/.agents/skills/qiaomu-novel-generator
  194. python3 scripts/validate_skill.py
  195. python3 /Users/joe/.codex/skills/.system/skill-creator/scripts/quick_validate.py .
  196. python3 scripts/evaluate_story.py examples/sample-01-wuxia-suspense.md --fail-on-warning
  197. python3 scripts/evaluate_story.py examples/sample-02-sci-fi-memory.md --fail-on-warning
  198. ```
  199. 包结构:
  200. ```text
  201. SKILL.md
  202. README.md
  203. manifest.json
  204. agents/interface.yaml
  205. agents/openai.yaml
  206. references/
  207. examples/
  208. scripts/
  209. assets/qiaomu-profile/
  210. ```
  211. ## Troubleshooting
  212. | 问题 | 常见原因 | 处理方式 |
  213. |---|---|---|
  214. | `npx skills add` 找不到 skill | GitHub repo 未更新、`SKILL.md` frontmatter 无效,或当前目录安装路径不对 | 先运行 `npx skills add joeseesun/qiaomu-novel-generator --list` |
  215. | 已安装但没有自动触发 | 当前 Agent 不支持隐式调用,或输入没有命中小说/故事触发词 | 明确说“用 qiaomu-novel-generator 写……” |
  216. | 故事变成设定说明 | 输入没有主角欲望、代价或危险 | 补一句“主角想要什么、会失去什么、谁在阻止他” |
  217. | 还没想清楚剧情 | 只有题材,没有冲突和结尾味道 | 让 skill 先给选项,回复 `按默认` 或 `1B 2A 3C` |
  218. | 套路太少 | 一直在同一题材里打转 | 要求“换一组剧情引擎”,例如悬疑惊、组织逆袭、黑色反转 |
  219. | 想借经典剧情 | 容易变成照搬 | 先让 skill 给“经典桥段启发”,选 2-3 个功能组合后再重构 |
  220. | 想先搜某部作品 | 容易复述原作 | 用 `source-research-remix` 只抽象功能、读者爽点和不可复制边界 |
  221. | 开篇看不懂 | 为了钩子使用了误导题材的隐喻 | 要求“开篇误读修复”,补上同句或下一句的现实锚点 |
  222. | AI 味重 | 公式化对比、教学腔、总结腔 | 要求“去 AI 味”,优先删改“不是X,而是Y” |
  223. <!-- qiaomu-profile:start -->
  224. ## 关于向阳乔木
  225. 向阳乔木(乔向阳 / Joe)是一位实践型 AI 产品与内容创作者,长期把前沿 AI 变化转译成可复用的工作流、产品判断、AI 编程实践、AI 搜索实践和 GEO/AI 营销方法。
  226. - 个人网站: https://qiaomu.ai
  227. - 博客: https://blog.qiaomu.ai
  228. - X: https://x.com/vista8
  229. - GitHub: https://github.com/joeseesun/
  230. - 微信公众号: 向阳乔木推荐看
  231. ### 支持与关注
  232. | 打赏支持 | 微信公众号 |
  233. |---|---|
  234. | <img src="assets/qiaomu-profile/qiaomu_reward_qr.png" alt="向阳乔木打赏二维码" width="180" /> | <img src="assets/qiaomu-profile/qiaomu_wechat_public_account_qr.jpg" alt="向阳乔木推荐看公众号二维码" width="180" /> |
  235. | 感谢支持乔木持续分享 AI 实践 | 扫码关注「向阳乔木推荐看」 |
  236. <!-- qiaomu-profile:end -->
  237. ## License
  238. MIT. See [LICENSE](LICENSE).
  239. ---
  240. <a name="english"></a>
  241. ## English
  242. `qiaomu-novel-generator` is an agent skill for writing original Chinese short fiction from a theme, character setup, trope, classic-story signal, or draft excerpt.
  243. It does not jump straight into generic prose. It first turns the idea into a story strategy: reader promise, pressure relationship, conflict arena, plot engines, outline, and ending aftertaste. Then it drafts the story and checks for hook strength, desire, escalation, dialogue pressure, scene imagery, reversal, ending resonance, anti-AI phrasing, and opening clarity.
  244. Install:
  245. ```bash
  246. npx skills add joeseesun/qiaomu-novel-generator
  247. ```
  248. Try:
  249. ```text
  250. Use qiaomu-novel-generator to write a complete wuxia suspense short story.
  251. ```
  252. ```text
  253. I want a story inspired by cultivation web novels. Give me the plot strategy first.
  254. ```
  255. ```text
  256. Search public references for this novel, extract the reusable craft functions, then remix them into a modern workplace power-struggle story.
  257. ```
  258. Core features:
  259. - option-based prewrite flow before drafting
  260. - classic beat remix without copying protected works
  261. - genre engines for wuxia, cultivation, suspense, romance, workplace, sci-fi, and short-drama-style hooks
  262. - anti-AI-language gate
  263. - opening-clarity gate to avoid misleading poetic hooks
  264. - sample evaluation scripts
  265. This skill does not copy protected text, famous scenes, signature lines, unique characters, or recognizable plot chains, and it does not directly imitate a living author's distinctive style.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.