Deai — Data skill for Claude Code
deai — de-AI writing for Chinese text.
How to install Deai
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open wei125775-lab/deai and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Deai does
deai — de-AI writing for Chinese text. One self-contained Agent Skill: strip AI flavor from web novels, fiction, articles, titles. Clean/Refactor/Detect/Write modes; layered 5-step pipeline; graded A/B/C rules; merges human-writing, lieflat, sepia, oh-story.
Alternatives in Data
- Archify — Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self 17.2k ★
- Autonovel — An autonomous novel writing pipeline, by Hermes Agent 1.5k ★
- Sprite Gen — Generate clean 2D game sprites & animation atlases — component-row pipeline: state rows, alpha cleanup, frame 763 ★
README
deai · 去 AI 味总刀
De-AI writing for Chinese text. A self-contained Agent Skill that strips AI flavor from web-novel prose, fiction, articles and titles — at the layer that actually gives AI away, not just word choice.
把"像 AI 写的"中文改成人写的。覆盖网文/小说正文、知乎回答、公众号、评论、标题。**自包含**:一个目录即一个 Agent Skill(仓库根是标准 SKILL.md),无运行时依赖。
它不是又一摞禁令,而是一条**纵深链 + 一套仲裁**:四家去 AI 味方法的规则被重写整合成一套带等级的单一流程,规则冲突按等级裁,避免各家规则互相打脸。
为什么跟别的去 AI 味工具不同
- 抓结构,不只抓词句。 AI 味最深藏在叙事架构(主题被角色亲口点破、每章恰好圆满、情绪全靠身体渲染、配角全对主角有好感),句子只是表层。结构层只在你授权"重构"时动刀,否则只标注。
- 分档操作。 清理(只清语言、剧情一字不动,默认)/重构(允许动结构骨架)/检测(只报告不改)/新写。
- 反误伤。 一条明确的 C 级禁动表:看着像 AI 味但实测站不住的东西(句长离散度、被动句、句内排比、问句、比喻本身),以及网文行业规范(强因果、伏笔全回收、主角主动),一律不碰。
- 校准哲学:对准人类带,不是反转 AI 特征。 每条规则都上满会造出新的指纹;架构层每篇只挑 2–5 个动作,句级命中即改但保持最小改动。
安装
仓库根是标准 Agent Skill(`SKILL.md` + `references/` + `scripts/`)。
- 手动复制(任意 agent):把整个目录复制到你的 skills 目录,如
~/.claude/skills/deai/。 - Skills CLI(77+ 个 agent):
npx skills add wei125775-lab/deai
装好后,对任意文本说「帮我去了 AI 味」「这段太 AI 了」「重写这段」即触发。不支持触发词时,显式问它「用 deai 处理这段」。
用法:四档
| 档 | 说 | 行为 |
|---|---|---|
| 清理(默认) | 「帮我去了 AI 味」 | 只清语言层,剧情结构一字不动;结构症状照标照提醒但不动刀 |
| 重构 | 「结构也能变」 | 允许动结构骨架;先给诊断报告 + 手术清单,授权后动刀 |
| 检测 | 「查下有没有 AI 味」 | 只报告,不改文 |
| 新写 | 「用去 AI 味的写法写」 | 从零写,落点即去 AI 味 |
怎么工作:纵深五步
| 步 | 层 | 读 | 档 |
|---|---|---|---|
| 0 | 预检:材料/信息守恒、说话位置、网文别误伤 | references/calibration.md |
全部 |
| 1 | 架构(虚构):结构诊断,挑 2-5 个动作 | references/architecture.md |
重构/新写 |
| 2 | 段落删剪:假深刻/注水/结尾早一拍 | references/passes.md |
清理+重构 |
| 3 | 句级门禁:A-G 触发标记,最小改动 | references/gates.md |
清理+重构 |
| 4 | 清零:硬禁词+标点,跑脚本量化 | references/hard-bans.md + scripts/check_deai.py |
清理+重构 |
| 5 | 冷读:忘掉规则重读,抓"像模型完成任务"的段 | references/passes.md 末节 |
改稿类 |
顺序铁律:先结构后词句,删除/重排优先于改写,脚本最后。结构有缺陷时只做句级修改,AI 味反而更显。
规则分级(仲裁层)
| 级 | 含义 | 动作 |
|---|---|---|
| A 硬信号 | 多源独立一致或实测稳定 | 命中必须改 |
| B 软信号 | 单源/需语境/分模型/作者风格 | 标出,由作者定,不擅动 |
| C 禁动 | 实测站不住或网文行业规范/作者风格 | 一律不碰 |
优先级:**作者风格文档 > C 禁动 > A 硬改 > B 软判**。例如破折号在不同后端模型上量级差 10 倍,等级按执行模型浮动(见 `refe
Related Skills
Fix Tickets
Autonomously fix one or more speckit-companion GitHub issues with the SpecKit Companion pipeline — clean branc
Jobs Add
Add a source to the pipeline in one of three modes — company (auto-detect ATS, default), board (enable a built
Great Writer
Universal AI writing skill — tech articles, marketing copy, research reports, Xiaohongshu notes, technical doc
AI Analyst Genome
AI Analyst Genome — self-contained bootstrap for building AI Data Analyst products with Claude Code. Powered b
Note To Artifact
Use when turning notes, data, or a plan into a beautiful, self-contained HTML artifact (report, dashboard, pla
Econ Paper Studio
End-to-end empirical economics workflow for AI agents Five-stage pipeline (question/design/execute/verify/writ
Related Agents
Repurpose Broadcast
Broadcast and publication content specialist. Generates WhatsApp Channel updates, Telegram Channel posts with
Science Agent
Improve scientific collaboration in AI-assisted research. Validate citations against CrossRef and BibTeX. Dete
Brain Eval Engineer
Evaluation engineer — question set tooling, layered metrics (harvest/graph/retrieval/answer), fixed-strategy a