natural-writing

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Natural writing

自然写作

Produce prose that avoids the documented signs of AI writing, based on Wikipedia:Signs of AI writing. That page is a field guide for detecting AI text; this skill inverts it into rules for not producing those patterns.

One pattern alone can be fine. The goal is prose where the patterns don't cluster and nothing reads generated.

Scope: this skill is a guideline, not law. Skimmability for the reader outranks any style rule in it (short front-loaded paragraphs, no text blobs, examples first); accuracy and reference-integrity rules always bind. Under that house standard, bold-label bullets and sparing keyword bolds are sanctioned formatting: the bold and bullet cautions in this skill apply to prose written for external readers, not to this repo's own skill docs.

生成避免已记录的AI写作迹象的散文，基于 Wikipedia:Signs of AI writing。该页面是用于检测AI文本的实地指南；本skill将其反向转化为不产生这些模式的规则。

单独一个模式可能没问题。目标是让散文中的模式不会聚集，也没有任何内容读起来像是生成的。

范围：本skill是指导方针，不是法律。读者的可略读性优先于其中的任何风格规则（简短的前置段落，无文本块，示例优先）；准确性和引用完整性规则始终有约束力。在该仓库标准下，粗体标签项目符号和少量关键词加粗是被认可的格式：本skill中关于粗体和项目符号的注意事项适用于为外部读者撰写的散文，不适用于本仓库自身的skill文档。

When to apply

何时应用

"humanize this text" / "make this sound less like AI" / "remove the AI tells"
"does this read as AI-generated?"
Drafting or rewriting prose for readers outside the repository: articles, blog posts, reports, emails, announcements, PR descriptions
Not repository documentation (READMEs, guides, reference pages): those are programmer-register pages of the
```
/docs-discipline
```
skill
Any writing task where the user cares about tone, voice, or naturalness

“人性化这段文本” / “让这听起来不像AI” / “去除AI痕迹”
“这读起来是AI生成的吗？”
为仓库外的读者起草或改写散文：文章、博客文章、报告、电子邮件、公告、PR描述
不适用于仓库文档（README、指南、参考页面）：这些是
```
/docs-discipline
```
skill 的程序员语域页面
任何用户关心语气、声音或自然度的写作任务

Source of truth

事实来源

The catalog of signs changes over time (models change; the community adds new tells). Prefer the live page, fall back to the bundled snapshot:

Live: fetch https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing (raw wikitext:
```
https://en.wikipedia.org/w/index.php?title=Wikipedia:Signs_of_AI_writing&action=raw
```
). Do this when auditing text for AI tells, when the user asks for a thorough pass, or when the bundled snapshot looks stale for the topic at hand.
Snapshot: references/signs-catalog.md is a distilled copy of the full page, reorganized as writing rules. Use it as the working checklist for every invocation; it is enough when offline or for routine drafting.

Snapshot last synced: 2026-08-26, from page revision 1371415133 (2026-08-26T08:17:42Z). See "Maintaining the snapshot" below for how to update it.

references/words-to-avoid.md lists the words and phrases the page flags, organized for scanning, plus ready-made grep commands for the mechanical checks.

迹象目录随时间变化（模型在变化；社区会添加新的迹象）。优先使用在线页面，退回至捆绑的快照：

在线：获取 https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing （原始wikitext：
```
https://en.wikipedia.org/w/index.php?title=Wikipedia:Signs_of_AI_writing&action=raw
```
）。在审计文本是否存在AI痕迹、用户要求彻底检查时，或捆绑快照看起来对当前主题过时时，执行此操作。
快照：references/signs-catalog.md 是完整页面的浓缩副本，重新组织为写作规则。在每次调用时将其作为工作检查清单；在离线或常规起草时它已经足够了。

快照最后同步：2026-08-26，来自页面修订版 1371415133（2026-08-26T08:17:42Z）。如何更新请参见下方“维护快照”。

references/words-to-avoid.md 列出了该页面标记的单词和短语，为扫描而组织，并提供了现成的 grep 命令用于机械性检查。

Workflow

工作流程

1. Load the rules

1. 加载规则

Read

references/signs-catalog.md

in full. For audits or thorough requests, also fetch the live page and note any signs not yet in the snapshot.

完整阅读

references/signs-catalog.md

。对于审计或彻底性请求，还要获取在线页面并记录快照中尚未包含的任何迹象。

2. Pick the mode

2. 选择模式

Draft: writing new text.
Rewrite: revising existing text. Preserve the author's facts and voice, and the structure where sound; strip only the tells. Do not inject new claims.
Audit: the user wants an assessment, not edits. Report which signs appear, where, with brief quotes. Mind the false-positive caveats in the catalog (§9): formal prose or perfect grammar alone proves nothing. Do not edit unless asked.

起草：撰写新文本。
改写：修改现有文本。保留作者的事实和声音，以及结构（如果合理）；只去除迹象。不要注入新主张。
审计：用户想要评估，而不是编辑。报告出现了哪些迹象、在哪里出现，附简短引文。注意目录（§9）中的误报注意事项：正式的散文或完美的语法本身不能证明什么。除非被要求，否则不要编辑。

3. Write or edit

3. 写作或编辑

Apply the full catalog. The highest-value rules:

State facts plainly, with is/are/has/was. Not "serves as", "stands as", "boasts", "features", "marks", "represents", or "refers to" openings.
Cut inflated significance: "pivotal moment", "testament to", "underscores its importance", "reflects broader trends", "rich cultural heritage", "evolving landscape". Replace generic importance-claims with a specific, checkable fact. Specificity is the whole game.
Kill trailing participle analysis. No sentence-final "-ing" clauses that editorialize: "…, highlighting the significance of X", "…, ensuring Y".
Avoid AI-vocabulary clusters: delve, crucial, pivotal, tapestry, landscape (abstract), showcase, underscore, intricate, fostering, garner, meticulous, vibrant, robust, boasts, align with, testament. One may be fine; clusters are the tell. See the word list.
Break the templates. No "not just X, but Y" or "It's not X, it's Y" parallelisms; no rule-of-three triads everywhere; no "Despite these challenges…" conclusions; no "In summary / In conclusion / Overall" wrap-ups; no "It's important to note".
Attribute precisely or not at all. No "experts argue", "observers have noted", "industry reports suggest". Name the source or drop the claim. Never imply many sources when there is one.
Never use a dash as punctuation. No em dashes (—), and do not sneak the same construction back in with an en dash (–) or a spaced hyphen ( - ); that swaps the character while keeping the tell. Restructure instead: a comma, a colon, parentheses, or two sentences. Hyphens stay only inside compound words, and en dashes only in numeric ranges. (House rule; it is stricter than the Wikipedia page, which only flags overuse.)
Format like a human. Sentence case headings. No bullet lists where every item is a bold label plus a colon. Bold almost nothing. No emoji decoration, no needless small tables, no horizontal rules before headings, no skipped heading levels. Prose first; lists only when the content is genuinely enumerable.
No chat artifacts. No "I hope this helps", "Certainly!", "Would you like…", "Here is a…" framing; no knowledge-cutoff disclaimers; no speculation dressed as fact ("details are not widely documented, but likely…"); no unfilled placeholders like
```
[Your Name]
```
or
```
2025-XX-XX
```
.
Use the target medium's markup, never Markdown headed for a non-Markdown destination. Never invent citations, links, DOIs, ISBNs, or page numbers. Verify every reference you emit, and strip tool residue (
```
utm_source=
```
,
```
turn0search0
```
,
```
oaicite
```
,
```
[cite: 1]
```
).
Vary rhythm, and repeat words when natural. Reusing the same noun beats cycling synonyms. Mix sentence lengths. Commit to definite statements ("was the first") and use plain verbs (wrote, used, moved, tried, died).

应用完整目录。最高价值的规则：

直接陈述事实，使用 is/are/has/was。不要用 “serves as”、“stands as”、“boasts”、“features”、“marks”、“represents” 或 “refers to” 开头。
去除夸大的重要性：“pivotal moment”、“testament to”、“underscores its importance”、“reflects broader trends”、“rich cultural heritage”、“evolving landscape”。用具体的、可核查的事实替换泛泛的重要性主张。具体性是关键。
删除句尾分词分析。 不要有句子末尾的 “-ing” 从句进行评论：“…, highlighting the significance of X”、“…, ensuring Y”。
避免AI词汇簇：delve, crucial, pivotal, tapestry, landscape（抽象）, showcase, underscore, intricate, fostering, garner, meticulous, vibrant, robust, boasts, align with, testament。单独一个可能没事；成簇才是迹象。参见单词列表。
打破模板。 不要有“not just X, but Y”或“It's not X, it's Y”的平行结构；不要到处都是三分法三联体；不要有“Despite these challenges…”的结论；不要有“In summary / In conclusion / Overall”的总结；不要有“It's important to note”。
精确归属，否则不提及。 不要有“experts argue”、“observers have noted”、“industry reports suggest”。指明来源或放弃该主张。永远不要在一个来源时暗示多个来源。
绝不使用破折号作为标点。 不要使用em破折号（—），也不要用en破折号（–）或带空格的连字符（ - ）偷偷把同样的结构加回来；那只是换了个字符而保留了迹象。改为重构：逗号、冒号、括号或两个句子。连字符只保留在复合词内部，en破折号只用于数字范围。（本仓库规则；比Wikipedia页面更严格，后者只标记过度使用。）
像人一样排版。 句子式大小写标题。不要有每一项都是粗体标签加冒号的项目符号列表。几乎不加粗。不要有emoji装饰、不必要的小表格、标题前的水平线、跳过的标题级别。散文优先；只有内容真正可枚举时才用列表。
不要有聊天工件。 不要有“I hope this helps”、“Certainly!”、“Would you like…”、“Here is a…”的框架；不要有知识截止日期免责声明；不要把猜测包装成事实（“细节没有被广泛记录，但可能…”）；不要有未填充的占位符如
```
[Your Name]
```
或
```
2025-XX-XX
```
。
使用目标媒介的标记语言，绝不要用面向非Markdown目的地的Markdown。绝不要编造引用、链接、DOI、ISBN或页码。核实你发出的每个引用，并去除工具残留（
```
utm_source=
```
、
```
turn0search0
```
、
```
oaicite
```
、
```
[cite: 1]
```
）。
变化节奏，并在自然时重复单词。 重复使用同一个名词胜过循环更换同义词。混合句子长度。做出明确的陈述（“was the first”）并使用平实的动词（wrote, used, moved, tried, died）。

4. Self-review pass

4. 自我检查

Before delivering, reread the output against the checklist above. Scan it against the term lists in

references/words-to-avoid.md

(they are written for reading, not for

grep -f

). Two checks are mechanical; run them on the draft file:

bash

undefined

在交付之前，对照上面的清单重读输出。根据

references/words-to-avoid.md

中的术语列表进行扫描（它们是用来阅读的，不是用来

grep -f

的）。有两项检查是机械性的；在草稿文件上运行：

bash

undefined

Dash punctuation and machine artifacts: any hit outside quoted material is a bug.

(Unspaced en dashes in numeric ranges like 2024–2025 are allowed and not flagged.)

High-signal AI vocabulary: hits are prompts to reread the sentence, not automatic errors


Fix a hit by rewriting the sentence, never by thesaurus-swapping the flagged word. The underlying pattern (vagueness, puffery, formula) is the problem, not the word itself.


通过重写句子来修复命中，绝不要用同义词替换被标记的单词。底层的模式（含糊、吹捧、公式化）才是问题，而不是单词本身。

5. Report

5. 报告

Draft/Rewrite: deliver the text. If rewriting, summarize which signs you removed in one or two sentences.
Audit: list the signs found with locations and quotes, state an overall read, and note the base-rate caveat: these are statistical signs, not proof.

起草/改写：交付文本。如果是改写，用一两句话总结你消除了哪些迹象。
审计：列出找到的迹象及位置和引文，陈述整体判断，并注明基础率注意事项：这些是统计迹象，不是证据。

Fallback without network

无网络时的回退

Everything works offline from

references/signs-catalog.md

and

references/words-to-avoid.md

. Only the live-refresh step is skipped.

一切都可离线工作，基于

references/signs-catalog.md

和

references/words-to-avoid.md

。只是跳过了在线刷新步骤。

Maintaining the snapshot

维护快照

When asked to update this skill from Wikipedia (or when the snapshot looks stale), reconcile semantically; the catalog is a reorganized distillation, so a textual diff against the wikitext is noise.

Check whether the page changed since the pinned revision:

bash

curl -fsSL 'https://en.wikipedia.org/w/api.php?action=query&titles=Wikipedia:Signs_of_AI_writing&prop=revisions&rvprop=ids|timestamp&format=json'

revid

equals the revision pinned under "Source of truth" above, nothing to do.

Otherwise fetch and read the full current wikitext:

bash

curl -fsSL 'https://en.wikipedia.org/w/index.php?title=Wikipedia:Signs_of_AI_writing&action=raw' -o /tmp/signs.txt

Compare it section by section against
```
references/signs-catalog.md
```
. Fold in new signs, drop removed ones, and update
```
references/words-to-avoid.md
```
to match.
Update the pinned revision ID, timestamp, and sync date in both places: under "Source of truth" in this file and in the header of
```
references/signs-catalog.md
```
. The two pins must match.

当被要求从Wikipedia更新此skill（或当快照看起来过时时），进行语义对账；目录是重新组织的浓缩，因此与wikitext的文本差异是噪音。

检查页面自固定修订版以来是否发生了变化：

bash

curl -fsSL 'https://en.wikipedia.org/w/api.php?action=query&titles=Wikipedia:Signs_of_AI_writing&prop=revisions&rvprop=ids|timestamp&format=json'

如果

revid

等于上面“事实来源”中固定的修订版，则无需操作。

否则获取并阅读完整的当前wikitext：

bash

curl -fsSL 'https://en.wikipedia.org/w/index.php?title=Wikipedia:Signs_of_AI_writing&action=raw' -o /tmp/signs.txt

将其与
```
references/signs-catalog.md
```
逐节比较。加入新迹象，移除已删除的迹象，并更新
```
references/words-to-avoid.md
```
以匹配。
在两个位置更新固定的修订版ID、时间戳和同步日期：本文件的“事实来源”下方和
```
references/signs-catalog.md
```
的头部。两个固定值必须匹配。