Agent skill

Fiction Distiller

by dama-cyber in dama-cyber/magic-distillation

短篇换元仿写器 - 从小说文本中提取叙事骨架,将人物/情节/设定/台词全部替换为原创内容,生成字数相近、风格神似、内容全新、可反检测的短篇小说

MITAuto-check passedWriting & Content

Install Fiction Distiller

skills CLI
$ npx skills add dama-cyber/magic-distillation --skill fiction-distiller -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install dama-cyber/magic-distillation fiction-distiller --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/dama-cyber/magic-distillation.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/fiction-distiller .claude/skills/fiction-distiller && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
fiction-distiller
GitHub stars
106
Token cost
~1.9k tokens
SKILL.md length
369 words
Files
82
Skills in repo
1
Repo updated
First seen
Licence
MIT

At a glance

短篇换元仿写器 - 从小说文本中提取叙事骨架,将人物/情节/设定/台词全部替换为原创内容,生成字数相近、风格神似、内容全新、可反检测的短篇小说

  • Tasks that involve Creative writing and fiction
  • SKILL.md covers 知识路径解析, 输出路径, 角色定位 and 懒加载协议, plus 4 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Fiction Distiller is an agent skill from dama-cyber/magic-distillation. 短篇换元仿写器 - 从小说文本中提取叙事骨架,将人物/情节/设定/台词全部替换为原创内容,生成字数相近、风格神似、内容全新、可反检测的短篇小说

Its SKILL.md is about 1.9k tokens, which your agent loads only when the skill is triggered. The skill folder holds 85 other files (for example `knowledge/ai-distillation/01-feature-extraction.md`, `knowledge/ai-distillation/02-fingerprint-verification.md` and `knowledge/ai-distillation/03-imitation-strategies.md`).

It sits in Writing & Content, covering Creative writing and fiction. The repository describes itself as: 稗官风骨全维蒸馏之器——自诸说部文字中提取格调之印,存其本味,别撰新篇。 The licence is MIT.

When your agent uses it

  • Tasks that involve Creative writing and fiction

Example prompts

  • “/fiction-distiller”

What it can do on your machine

Read from SKILL.md and the folder at commit 1863288. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Fiction Distiller loads about 1.9k tokens when it runs. Until then it costs about 22 tokens; SKILL.md has 369 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~22
When it runs · the whole SKILL.md, loaded when a task matches
~1.9k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from dama-cyber/magic-distillation at commit 1863288, republished under its MIT licence (© dama-cyber). 369 words, ~1,886 tokens.

Download SKILL.mdSave it as .claude/skills/fiction-distiller/SKILL.md (or your agent's skills folder). This skill also uses 81 other files; get the full folder from GitHub.
name
fiction-distiller
description
短篇换元仿写器 - 从小说文本中提取叙事骨架,将人物/情节/设定/台词全部替换为原创内容,生成字数相近、风格神似、内容全新、可反检测的短篇小说
metadata.tier
base
metadata.type
fiction-writing
metadata.version
4.1

fiction-distiller

短篇换元仿写技能 | 版本 4.1(核心修复版)

知识路径解析

本技能中 @skill:knowledge/... 引用指向 SKILL.md 同级的 knowledge/ 目录。映射规则:

@skill:knowledge/index.md              → knowledge/index.md
@skill:knowledge/general-style/01-stylistics → knowledge/general-style/01-stylistics.md

所有 @skill:knowledge/X/Y 均对应 knowledge/X/Y.md(自动追加 .md 扩展名)。加载时直接读取对应路径。

输出路径

所有生成的内容统一保存到 SKILL.md 同级的 output/ 目录下。按类型和章节逐文件保存,一章一文件。

严禁修改或覆盖 output/ 以外的任何项目文件(SKILL.md、knowledge/、CLAUDE.md、README.md 等)。只读读取,只写 output。

内容类型命名规则示例
风格模型卡风格模型卡-《作品名》.md风格模型卡-《百年孤独》.md
长篇正文第N章-《故事名》.md第01章-《雾港》.md
长篇大纲大纲-《故事名》.md大纲-《雾港》.md
短篇正文《故事名》.md《回声》.md
分析报告分析-《作品名》.md分析-《局外人》.md

角色定位

你是世界顶尖的文学风格分析师与创意写作大师。你拥有深厚的文学理论功底、敏锐的风格感知力和精湛的创作技艺。你的核心能力是:从任何小说文本中蒸馏出风格精华,并以该风格创作全新的原创故事。

懒加载协议

禁止一次性加载全部知识文件。严格按三层协议分步加载,每一步完成后再触发生下一步:

[触发条件]              [加载内容]             [完成后]
步骤 1(知识索引)   → index.md 摘要       → 触发步骤 2
步骤 2(文本接收后) → 基础层(始终加载)   → 触发步骤 3
步骤 3(蒸馏开始)   → 体裁层(按类型加载) → 触发步骤 4-5
步骤 4-5(创作阶段) → 方法层(按需加载)   → 触发步骤 6
步骤触发时机加载范围文件
① 索引始终首先执行仅 index.md 摘要@skill:knowledge/index.md
② 基础层步骤 2 文本接收后通用分析 + 指纹方法论(11 个文件)general-style/* + author-fingerprint/*
③ 体裁层步骤 3 蒸馏开始时按文本类型选一组(5-7 个文件)long-fiction/* / short-fiction/* / net-style/*
④ 方法层步骤 4-5 创作阶段创作方法 + 验证(4 个文件)ai-distillation/*

全自动执行流程

步骤 1:知识索引构建

读取 @skill:knowledge/index.md,按三层优先级构建分析矩阵:

  • 基础层(始终首先加载摘要):general-style/* + author-fingerprint/*
  • 体裁层(按用户文本类型选择性深入加载):long-fiction/* / short-fiction/* / net-style/*
  • 方法层(创作阶段加载):ai-distillation/*
步骤 2:文本接收与智能切分
文本类型自动处理方式
长篇(>5万字)按章节标题/分卷标记/逻辑节点切分 → 逐章独立蒸馏 → 汇总跨章节规律 → 生成全书风格模型卡
短篇(<5万字)直接全文蒸馏
多篇文本逐篇蒸馏 → 对比分析 → 提取跨作品风格核心
步骤 3:全维度风格蒸馏

从以下 11 个维度逐一解构,汇总为风格模型卡:

#维度分析重点参考知识文件
1叙事视角与声音叙述者类型、人格显现方式、可靠性@skill:knowledge/author-fingerprint/04-narrative-voice
2词汇与用语指纹高频特征词、词性偏好、自创词/方言@skill:knowledge/author-fingerprint/01-lexical
3句式工程与节奏句长分布、长短句切换规律、非常规句法@skill:knowledge/author-fingerprint/02-syntactic
4标点与排版指纹特殊标点用法、视觉节奏设计@skill:knowledge/author-fingerprint/05-punctuation
5感官描写与五感权重主导感官、通感运用、感官切换节奏@skill:knowledge/general-style/05-imaging
6人物构建与对话风格人物功能、对话特征、人物声音差异化@skill:knowledge/net-style/04-characters
7情节结构与时空操控冲突模型、时间处理、空间调度@skill:knowledge/long-fiction/01-story-layer
8修辞系统与意象谱系比喻来源域、核心意象、象征系统@skill:knowledge/author-fingerprint/03-rhetorical
9氛围底色与情感色调主导情绪、色调一致性、情感节奏@skill:knowledge/net-style/06-narrative-focus
10主题领域与哲学立场核心命题、价值取向、世界观底色@skill:knowledge/general-style/03-rhetoric
11网文特殊维度(按需)爽点类型/密度、金手指、升级节奏、世界观设定@skill:knowledge/net-style/02-pleasure-mechanism

风格模型卡格式:

【风格模型卡】
▸ 核心风格标签(5-8个,精准、可操作)
▸ 15条必遵风格规则(每条可量化、可自检)
▸ 风格指纹小结(300字,高度凝练)
▸ 反面清单(写作时绝对禁止的5个行为)
步骤 4:原创故事构思
类型设计内容目标篇幅
短篇独立人物 + 全新背景 + 单一核心冲突≈10,000 字
长篇完整人物群像 + 多卷弧线 + 世界观体系 + 分卷大纲 + 冲突演进框架≥500,000 字

原创隔离原则:情节、人物、对白、场景、专名与原文完全隔离,仅继承风格模型卡中的规则。

步骤 5:原创模仿创作

以"我就是原作者在写一个全新故事"的内在视角创作,写作时动态对标每一条风格规则:

类型输出内容
短篇完整正文(≈10,000 字)
长篇详细分卷大纲 + 前三章完整正文 + 后续章节详纲
步骤 6:原创性自检
自检项目:
□ 无任何连续 10 字以上与原文相同的句子
□ 核心情节与人物关系无实质雷同
□ 无沿用原文专有名词
□ 风格规则逐条验证通过

作品末尾附加:

【原创性声明】
本故事所有人物、情节、对白均为全新创作,不存在任何对原作的
直接引用或改写。模仿仅限于风格维度(叙事节奏、句式特征、修
辞偏好、感官侧重等),不涉及内容层面的借用。

v4.1 核心修复说明

v4.0 端到端测试发现三大核心问题,v4.1 已全部修复:

问题1:生成器本质是"换名微调"而非"功能等价重写"
  • 根因:生成 prompt 未明确禁止逐句改写,导致新篇与原文句式一一对应
  • 修复:p04-paragraph-generate.md 增加功能等价重写原则,增加"连续4字禁止"硬约束,增加正确/错误示例对比
  • 修复:p02-variable-annotate.md 增加影响层级、同义替换禁忌、原文危险片段清单
  • 修复:p03-new-variable-design.md 增加零残留标准从8字提至4字,禁止近义变体
Show full SKILL.md (149 more words)Show less
问题2:工具误报严重
  • 根因:originality_checker 4-6字匹配过于宽泛,anti_detection_checker 黑名单包含日常高频词
  • 修复:originality_checker 增加三级残留检测(真实残留/7字+潜在残留/4-6字日常短语),分离统计
  • 修复:anti_detection_checker 增加硬/软黑名单分级,软黑名单设合理阈值(≤5),模式化标记设阈值(≤3)
  • 修复:anti_detection_checker 增加短句占比和长短句交替率指标
  • 修复:两工具均增加上下文提取和段落级分析
问题3:字数/原创性/反检测未联动
  • 修复:p04 生成 prompt 已嵌入原创性约束和反检测约束
  • 修复:p05 人味化 prompt 增加原文危险片段扫描步骤
  • 修复:p06 字数修正 prompt 增加原创性约束(扩写时不得引入原文片段)
  • 新增:quality_report.py 统一质量报告工具,集成三维度检查

绝对红线

❌ 跳过知识索引
❌ 复制原文任何连续句子(≥4字)
❌ 沿用原文情节/场景/对白/专名
❌ 逐句对应原文改写(换名微调)
❌ 使用同义替换禁忌清单中的词
❌ 长篇以整体印象代替逐章分析
❌ 搜索未完成就开始编写文件
❌ 虚构搜索结果(所有内容必须来自真实搜索)
❌ 修改或覆盖 `output/` 外的任何项目文件(只读项目,只写 output)

知识文件引用索引(含层级标记)

加载协议:步骤 ① 读 index.md → 步骤 ② 读 🔴 基础层(共 11 个) → 步骤 ③ 选读 🟡 体裁层(按文本类型 5-7 个) → 步骤 ④-⑤ 读 🟢 方法层(共 4 个)

🔴 基础层(步骤 ② 始终加载,共 11 个文件)
文件核心内容
@skill:knowledge/general-style/01-stylistics文学文体学:语域/句法/词汇/修辞/音韵
@skill:knowledge/general-style/02-narratology叙事学:叙述者分层与叙事时间
@skill:knowledge/general-style/03-rhetoric修辞学:象征行动与隐喻系统
@skill:knowledge/general-style/04-stylometry文体计量学:词频统计与聚类分析
@skill:knowledge/general-style/05-imaging语言成像理论:名词独立句的成像作用
@skill:knowledge/author-fingerprint/01-lexical词汇指纹:高频特征词与词性偏好
@skill:knowledge/author-fingerprint/02-syntactic句式指纹:句长分布与长短句节奏
@skill:knowledge/author-fingerprint/03-rhetorical修辞指纹:比喻体系与通感映射
@skill:knowledge/author-fingerprint/04-narrative-voice叙事声音指纹:叙述者人格与可靠性
@skill:knowledge/author-fingerprint/05-punctuation标点指纹:破折号/省略号/引号用法
@skill:knowledge/author-fingerprint/06-cross-work跨作品一致性:风格核心与演变轨迹
🟡 体裁层(步骤 ③ 按文本类型选一组加载,共 17 个文件)

网文类(用户文本含网文特征时加载,7 个文件):

文件核心内容
@skill:knowledge/net-style/01-genres网文流派与类型体系
@skill:knowledge/net-style/02-pleasure-mechanism快感机制:爽点分类与心理学基础
@skill:knowledge/net-style/03-world-building世界观设定:四种模式与三层架构
@skill:knowledge/net-style/04-characters人设与角色:功能角色与金手指
@skill:knowledge/net-style/05-rhythm-structure节奏与结构:黄金三章与断章技巧
@skill:knowledge/net-style/06-narrative-focus叙事重心:事业线/感情线与情感策略
@skill:knowledge/net-style/07-era-features网文时代特征:2000s→2020s演变

长篇文学类(用户文本为 5 万字以上纯文学时加载,5 个文件):

文件核心内容
@skill:knowledge/long-fiction/01-story-layer长篇故事层面:冲突模型与编排
@skill:knowledge/long-fiction/02-narrative-layer长篇叙述层面:叙述者与聚焦模式
@skill:knowledge/long-fiction/03-structure长篇结构设计:起承转合与多线结构
@skill:knowledge/long-fiction/04-style长篇文体层面:五感权重与隐喻系统
@skill:knowledge/long-fiction/05-revision长篇修改打磨:删减与强化技术

短篇文学类(用户文本为 5 万字以下纯文学时加载,5 个文件):

文件核心内容
@skill:knowledge/short-fiction/01-elements短篇唯一效果原则:爱伦坡效应
@skill:knowledge/short-fiction/02-strategic-choice短篇战略选择:契诃夫细节法
@skill:knowledge/short-fiction/03-details短篇细节真实:选择标准与密度控制
@skill:knowledge/short-fiction/04-structure短篇结构设计:欲望+障碍+行动
@skill:knowledge/short-fiction/05-three-beauties短篇美学边界:自由/想象/超越
🟢 方法层(步骤 ④-⑤ 创作阶段加载,共 4 个文件)
文件核心内容
@skill:knowledge/ai-distillation/01-feature-extraction文本特征提取:词频/嵌入/句法树
@skill:knowledge/ai-distillation/02-fingerprint-verification风格指纹验证:作者识别算法
@skill:knowledge/ai-distillation/03-imitation-strategies风格模仿策略:提示工程与约束生成
@skill:knowledge/ai-distillation/04-evaluation原创性评估:分类器评估与人类盲测

© dama-cyber, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 81 other files in .claude/skills/fiction-distiller of dama-cyber/magic-distillation.

  • SKILL.md
  • knowledge/ai-distillation/01-feature-extraction.md
  • knowledge/ai-distillation/02-fingerprint-verification.md
  • knowledge/ai-distillation/03-imitation-strategies.md
  • knowledge/ai-distillation/04-evaluation.md
  • knowledge/anti-detection-short/01-detection-principles.md
  • knowledge/anti-detection-short/02-humanization.md
  • knowledge/anti-detection-short/03-rhythm-perturbation.md
  • knowledge/anti-detection-short/04-ai-word-list.md
  • knowledge/anti-detection-short/05-short-verification.md
  • knowledge/author-fingerprint/01-lexical.md
  • knowledge/author-fingerprint/02-syntactic.md
  • knowledge/author-fingerprint/03-rhetorical.md
  • knowledge/author-fingerprint/04-narrative-voice.md
  • knowledge/author-fingerprint/05-punctuation.md
  • knowledge/author-fingerprint/06-cross-work.md
  • knowledge/general-style
  • … and 65 more

Open the folder on GitHubat commit 1863288

Compare with similar skills

Fiction Distiller next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Fiction Distiller compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Fiction Distiller this skilldama-cyber/magic-distillation106—~1.9kAutomated safety check: PassMIT
Story Multi-Perspective Reviewzenstory-ai/oh-story-claudecode7.4k3 repos~3kAutomated safety check: PassMIT
Short Web Fiction Trend Scanzenstory-ai/oh-story-claudecode7.4k2 repos~1.2kAutomated safety check: PassMIT
InkOS Creative HarnessNarcooo/inkos10k1 repos~1.1kAutomated safety check: PassAGPL-3.0
Novel Arteternityspring/shuohao-skills4.3k—~1.1kAutomated safety check: NotesApache-2.0
SepiaNanako0129/sepia3.1k—~3.6kAutomated safety check: PassMIT

Similar skills

  • Story Multi-Perspective Review

    zenstory-ai/oh-story-claudecode

    Reviews Chinese web-novel text with several reviewer agents in parallel, falling back to a single-agent pass, and reports structure, character, prose and setting problems with fixes.

    7.4k GitHub starsUsed in 3 repos~3k tokens
    Writing & ContentAuto-check passed
  • Short Web Fiction Trend Scan

    zenstory-ai/oh-story-claudecode

    Scans popular short web-fiction rankings on Chinese platforms such as Dianzhong and Heiyan to surface trending emotional hooks, themes and topic candidates with an expiry warning.

    7.4k GitHub starsUsed in 2 repos~1.2k tokens
    Writing & ContentAuto-check passed
  • Drives long-form fiction, scripts, storyboards, interactive films and long-document translation through InkOS, with every change made by a typed action.

    10k GitHub starsUsed in 1 repo~1.1k tokens
    Writing & ContentAuto-check passed
  • Novel Art

    eternityspring/shuohao-skills

    给 AI 短剧出美术设定集(场景 + 叙事道具):场景的设计意图、一致性锚点、光照时段变体、 空景提示词;道具的戏剧功能、状态变体、尺度参照、白底无手提示词。

    4.3k GitHub stars~1.1k tokensUpdated 3 days ago
    Writing & ContentAuto-check: notes
  • Sepia

    Nanako0129/sepia

    Make AI-generated writing read as human-written, in fiction and in professional prose.

    3.1k GitHub stars~3.6k tokensUpdated yesterday
    Writing & ContentAuto-check passed
  • Story Toolbox Router

    zenstory-ai/oh-story-claudecode

    Routes a Chinese web-novel writing request to the matching tool in a 13-skill toolbox, covers author habit memory, and can launch a local dashboard for browsing a project.

    7.4k GitHub stars~1.7k tokensUpdated yesterday
    Writing & ContentAuto-check passed

Questions about Fiction Distiller

What does Fiction Distiller do?

短篇换元仿写器 - 从小说文本中提取叙事骨架,将人物/情节/设定/台词全部替换为原创内容,生成字数相近、风格神似、内容全新、可反检测的短篇小说. Fiction Distiller is an agent skill from dama-cyber/magic-distillation.

When should I use Fiction Distiller?

Fiction Distiller fits situations like: tasks that involve Creative writing and fiction.

How do I install Fiction Distiller in Claude Code?

Run `npx skills add dama-cyber/magic-distillation --skill fiction-distiller -a claude-code`. Or copy the skill folder (.claude/skills/fiction-distiller in dama-cyber/magic-distillation) into .claude/skills/fiction-distiller in your project. Claude Code loads it when a task matches its description.

How do I install Fiction Distiller in Codex?

Run `npx skills add dama-cyber/magic-distillation --skill fiction-distiller -a codex`. Or copy the skill folder (.claude/skills/fiction-distiller in dama-cyber/magic-distillation) into .agents/skills/fiction-distiller in your project. Codex loads it when a task matches its description.

Can I use Fiction Distiller in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add dama-cyber/magic-distillation --skill fiction-distiller -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/fiction-distiller, .gemini/skills/fiction-distiller, .github/skills/fiction-distiller and .opencode/skills/fiction-distiller in your project.

What does Fiction Distiller need to run?

SKILL.md names no scripts, command-line tools or credentials: Fiction Distiller is instructions for the agent only.

Does Fiction Distiller access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Fiction Distiller safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Fiction Distiller use?

Fiction Distiller is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Fiction Distiller use?

About 1.9k tokens (SKILL.md is roughly 7.5k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Fiction Distiller?

Skills that share tags, products or a category with Fiction Distiller: Story Multi-Perspective Review (zenstory-ai/oh-story-claudecode, 7.4k stars), Short Web Fiction Trend Scan (zenstory-ai/oh-story-claudecode, 7.4k stars), InkOS Creative Harness (Narcooo/inkos, 10k stars) and Novel Art (eternityspring/shuohao-skills, 4.3k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Fiction Distiller?

dama-cyber (a GitHub user) maintains it in dama-cyber/magic-distillation, which has 106 GitHub stars. The repository was last updated on June 18, 2026.

Source: dama-cyber/magic-distillation on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.