agentleFS
BooleSystemsBooleSystems/BooleOS/.claude/skills/de-ai-writing/SKILL.md

de-ai-writing

Skill2 starsChanged 48 days agoMITView on GitHub

Rewrite text so it has no signs of AI writing, using Wikipedia's "Signs of AI writing" field guide (em-dash overuse, "not just X, it's Y", "serves as", "delve/crucial/pivotal/tapestry", puffed-up significance, promotional tone, vague attributions, "it's important to note", "In conclusion", Title Case headings, bold-label bullet lists, chatbot leftovers). Use this whenever someone wants content de-AI'd, humanised, made to "sound less like ChatGPT/AI", "less robotic", cleaned of AI tells, or checked for signs of AI writing before publishing, including blog posts, web copy, FAQs, emails, essays, LinkedIn posts, product pages, docs and whole websites or repos of content. Also use it when polishing a draft the user says came from an AI.

7 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use good-writing.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Ayiming12Skill

    good-writing

    复现特定作者中文写作风格的轻量 Skill。用于从零写作、改写、终稿精修、风格审阅和结构保真翻译。默认先读风格摘要,不再启动时通读完整范文或大型检测清单。

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Scan it: If you can run code, use the bundled scanner for a baseline:

    line 28
  2. Distinctive:

    Unknowns: if the original says something isn't known, keep that honesty, but say it once, concretely, with what the reader should do …

    line 49
  3. Distinctive:

    Re-scan and re-read: Run the scanner again and aim for zero hits.

    line 56

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

De-AI writing

Rewrite content so it reads as written by a knowledgeable person, by removing the patterns catalogued in Wikipedia's Signs of AI writing (WikiProject AI Cleanup). The full catalogue, with words to watch and a before/after fix for each sign, is in references/signs.md. Read it before the first rewrite in a session; it is the source of truth for what counts as a sign.

Why this works the way it does

The Wikipedia guide makes a point worth holding on to: the signs are symptoms. Language models regress to the most statistically common phrasing, so specific facts get smoothed into generic, important-sounding statements. Swapping a flagged word for a synonym, or an em dash for an en dash, hides the symptom and keeps the problem. The aim is text that says specific things plainly. When you fix a sentence, fix what it is doing (puffing up, hedging, correcting a misconception nobody raised), not just the word that tripped the scanner.

Workflow

  1. Get the text. Pasted text, a file, a folder of files, or a built website. For a codebase or site, find where the copy actually lives (data files, CMS exports, templates) and edit there, not in generated output.
  2. Scan it. If you can run code, use the bundled scanner for a baseline:
    python3 scripts/check_ai_signs.py <file(s)>      # .txt, .md or .html; "-" reads stdin
    python3 scripts/check_ai_signs.py <file> --summary
    It reports each matching sentence under its sign number. It catches pattern-matchable signs only. Rule of three, elegant variation, puffed-up structure and summary endings need your own read. Without code execution, do the same pass by hand using references/signs.md.
  3. Rewrite. Go sentence by sentence through everything flagged, then read the whole piece once more for the judgement signs. Rules for the rewrite:
    • Keep every fact, number, name, date, link and claim. De-AI'ing changes how things are said, never what is said. If a sentence was only padding, delete it; if it carried a fact, keep the fact.
    • Keep the author's voice and conventions: person (I/we/you), spelling (UK/US), tone, formatting system (Markdown, HTML, JSX, plain text), and any length limits (SEO titles ~60 characters, meta descriptions ~155, button labels).
    • Never alter quotations of real people or sources, titles of works, or legal text.
    • Em dashes: rewrite with commas, colons, parentheses or full stops. Don't substitute another dash. One or two in a long piece is fine.
    • Contrasts ("not X but Y", "rather than", "not just… it's"): state the positive claim on its own unless a reader genuinely would assume the opposite.
    • Unknowns: if the original says something isn't known, keep that honesty, but say it once, concretely, with what the reader should do ("Hours aren't posted; call ahead"). Don't pad it with speculation.
    • Plain words: is/has instead of serves as/boasts; used, tried, moved, to, before.
    • Don't over-correct. Perfect grammar, formal register, a single transition word or an unsourced claim are not signs (see section 7 of the reference). Leave clean sentences alone.
  4. Re-scan and re-read. Run the scanner again and aim for zero hits. A hit you keep on purpose (a quotation, a proper noun like "Key Largo", a word used literally, such as "landscape" for a painting) is fine; note it in your report.
  5. Report briefly. Give the rewritten text (or confirm files were edited), then a short note: sign counts before → after, two or three representative before/after examples, and anything left in deliberately. Keep the note short; the rewrite is the deliverable.

Quick reference

The signs, in brief (details and fixes in references/signs.md):

#SignTypical tells
1.1Undue significance / legacystands as, testament, pivotal role, reflects broader, enduring legacy
1.2Canned notabilityfeatured in, independent coverage, active social media presence
1.3Superficial -ing analysis"…, highlighting/ensuring/reflecting…", valuable insights
1.4Promotional tonevibrant, nestled, in the heart of, boasts, world-class, seamless
1.5Vague attributionexperts say, observers note, studies show, widely regarded
1.6Challenges formula"Despite its…, faces challenges… Despite these challenges…"
1.7Title-defining lead"X refers to…"
2.1AI vocabularyadditionally, crucial, delve, enhance, foster, key, landscape, robust, tapestry
2.2Copula avoidanceserves as, functions as, represents, features/offers (for "has")
2.3Vague associationassociated with, in connection with
2.4Negative parallelismnot only…but, it's not X it's Y, no X no Y just Z, rather than
2.5Rule of threestacked triplets used for rhythm
2.6Elegant variationsynonym cycling for one noun
3.1–3.8FormattingTitle Case headings, bold everywhere, Label: bullets, em dashes, emoji, tiny tables, mixed curly quotes, --- between sections
4.1–4.3Chat leftoversI hope this helps, let me know, "not widely documented", [Your Name]
5.1–5.2Older habitsit's important to note, may vary, In conclusion, Overall

Example

Input

Nestled in the heart of the city, our café stands as a testament to the town's rich cultural heritage — it's not just a coffee shop, it's a gathering place. Additionally, the menu boasts a diverse array of pastries, showcasing our commitment to quality. While specific details about opening hours are not widely documented, it's important to note that hours may vary.

Output

Our café is on the main square, in a building that has been a bakery since 1911. People come for the coffee and stay for hours. We bake about twenty kinds of pastry each morning. Opening hours change with the seasons, so check our Instagram before you visit.

(The details "1911", "twenty kinds" and "Instagram" would come from the user. If the facts aren't available, ask, or write the plain version without them: "Our café is on the main square. We bake pastries every morning. Hours change with the seasons, so check before you visit." Never invent facts to replace puffery.)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.