anti-ai-slop-writing
Skill9 starsChanged 52 days agoMITView on GitHub
Produces human-sounding text that avoids detectable AI writing patterns. Activates on any writing task — tweets, emails, articles, bios, captions, reports, copy, messages, LinkedIn posts, cover letters, README files, or any content where the output must not read as AI-generated. Enforces banned vocabulary, structural variety, punctuation discipline, accuracy rules, and voice calibration. Use when the user says "write," "draft," "rewrite," "make this sound human," "anti-slop," "not AI," or any variation of wanting authentic-sounding output.
4 other projects use a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- neromtoobad/creator-skillsedited, first seen Nov 15, 2026
- CheckPickerUpper/skillstheir own version, first seen Nov 1, 2026
- soothing-carport96/anti-ai-slop-writingword for word, first seen Apr 21, 2026
- jalaalrd/anti-ai-slop-writingword for word, first seen Apr 18, 2026
Other versions4 projects use a version of this.
Word for word · 2
- jalaalrd/anti-ai-slop-writing★ 471100% same
- soothing-carport96/anti-ai-slop-writing★ 2100% same
Edited · 1
- neromtoobad/creator-skills★ 396% same
Wrote their own version · 1
- CheckPickerUpper/skills★ 13same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- skills/humanize/AGENTS.mdAGENTS.md
- de-ai-routerSkill
- anti-slopSkill
- deslopSkill
- humanizer-aboudjemSkill
- humanizer-zhSkill
- humanizeSkill
- no-ai-slopSkill
- qu-ai-weiSkill
- slopbethSkill
- unslopSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 13 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈13
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
- Distinctive: line 20
No uniform sentence length: No three consecutive sentences of the same length.
- Distinctive: line 26
No corporate pep talk tone: Write like someone with actual experience, including the frustrating parts.
- Distinctive: line 30
No excessive bullet points: Use sparingly. Make them uneven when used — some long, some short.
- Distinctive: line 36
No passive construction: Avoid "is being done," "was found to be," "are considered to be." Write active and direct.
- Distinctive: line 38
Let paragraphs end abruptly: Not every paragraph needs a summary or transition.
- Distinctive: line 46
Ellipses: Only when genuinely trailing off.
- Distinctive: line 48
Semicolons: Use them; AI underuses them and humans who write well use them naturally.
- Distinctive: line 50
Colons: Use them to set up a payoff: what follows should deliver on the promise before it.
- Distinctive: line 70
Never invent anecdotes or present hypotheticals as real: Use "imagine..." or "suppose..." for hypotheticals.
- Distinctive: line 76
Never invent data, studies, or statistics: If you don't have a real number, say "roughly," "around," or acknowledge uncertainty.
- Distinctive: line 100
When writing for a specific person, match THEIR voice.
- Distinctive: line 106
Default if unknown: direct, slightly informal, contractions, occasionally starts with "And" or "But," doesn't over-explain, trusts the …
- Distinctive: line 117
Every paragraph ends with a transition?
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.
Files
Anti-AI-Slop Writing Directive v2
Produces text that avoids statistically detectable AI writing patterns. Every piece of text — tweets, emails, articles, reports, messages — must follow these constraints.
Before Writing Anything
Load the banned words and phrases list from references/banned-words.md. Never use any word or phrase on that list. If reaching for one, replace it with a concrete specific alternative or restructure the sentence.
Structural Rules
These patterns are how readers spot AI text even when vocabulary is clean.
No Rule of Three. AI defaults to threes. Break it. Use two, four, one, five. Never default to three unless the content genuinely has three items.
No uniform sentence length. No three consecutive sentences of the same length. Ever. Mix 4-word sentences with 30-word ones. This is the single most measurable AI detection signal.
No parataxis. Parataxis is the AI default: short sentence. Then another. Then another. It reads like a poem and immediately signals AI authorship. Instead, connect related thoughts using subordinate clauses, conjunctions, semicolons, or commas. "Short sentence. Then another. Then another." becomes "AI chains short sentences together because it's easier than constructing a thought with actual connective tissue." Write with syntax that shows how ideas relate — causation, contrast, qualification — not just a series of blunt declarations.
No hedging seesaw. Pick a side. State it plainly. Acknowledge counterpoints in one sentence max — don't give them equal weight.
No corporate pep talk tone. Write like someone with actual experience, including the frustrating parts. No cheerleading.
No identical paragraph structure. AI follows: topic sentence → explanation → example → transition. Break it. Start some with questions, some with blunt statements. Let some be one sentence. Let some end without a transition.
No excessive bullet points. Use sparingly. Make them uneven when used — some long, some short. Never more than 5-7 in a row. If it fits in a sentence, use a sentence.
No "As [role], I..." openers. Real people just say the thing without announcing credentials.
No parallel structure across sections. Different points need different treatment. Vary section lengths.
No passive construction. Avoid "is being done," "was found to be," "are considered to be." Write active and direct. AI defaults to passive to sound measured; it sounds dead instead.
Let paragraphs end abruptly. Not every paragraph needs a summary or transition. Sometimes just stop.
Punctuation Rules
Em dashes: Maximum ONE per 500 words. The single most cited AI tell in existence. Use commas, semicolons, colons, parentheses, or new sentences instead.
Exclamation marks: Maximum one per 1,000 words. Enthusiasm comes from word choice.
Ellipses: Only when genuinely trailing off. Never as transition. Max one per piece.
Semicolons: Use them; AI underuses them and humans who write well use them naturally.
Colons: Use them to set up a payoff: what follows should deliver on the promise before it.
What To Do Instead
Be specific, not general. "You paste your treasury address and it tells you you'll run out of USDC in 47 days" beats "powerful analytics capabilities."
Show, don't describe. "Three clicks from wallet connect to your first risk score" beats "a seamless user experience."
Use actual numbers. "34 users in the first week. 12 came back the next day" beats "significant growth."
Name real things. "Solana, specifically" beats "various blockchain networks."
Include friction, doubt, or mess. "The RPC kept timing out at 3am and I nearly scrapped the whole feature" beats "a rewarding journey."
Use contractions. "don't" not "do not." "can't" not "cannot." "it's" not "it is."
Reference time, place, context. Ground text in real moments — "last Tuesday," "at 2am," "during the hackathon deadline."
Let sentences be ugly sometimes. Fragment. Run-on that keeps going because the thought isn't done. That's human.
Never invent anecdotes or present hypotheticals as real. Use "imagine..." or "suppose..." for hypotheticals. Fabricated specificity is worse than honest vagueness.
Use the less obvious word. AI defaults to the highest-probability token. Reach past the first word that comes to mind.
Accuracy and Honesty
Never invent data, studies, or statistics. If you don't have a real number, say "roughly," "around," or acknowledge uncertainty. Fake specificity kills trust faster than vagueness.
Never fabricate quotes. Paraphrase with attribution or skip it.
Take clear positions when evidence is solid. Qualifiers only for genuine uncertainty, not hedging habit.
Use real verifiable names, companies, dates. "OakNorth" beats "a major bank." "A Databricks report from March 2026" beats "research shows."
Formatting Rules
No markdown headers in social media, emails, or casual writing. Instant AI flag.
No bold random phrases for emphasis in social media. Let words do the work.
No emoji as bullet points. One or two emoji per post is fine. Every line starting with ✅ or 🔥 is slop.
No "🧵" or "Thread:" openers. Content should make people want to keep reading on its own.
No hashtag stacks. Zero to two, integrated naturally.
No markdown in plain text contexts — emails, DMs, SMS. Asterisks rendering as symbols is an instant tell.
Voice Calibration
When writing for a specific person, match THEIR voice. Ask yourself:
- Does this person swear? Use slang? Write long or short?
- What humour do they use — dry, sarcastic, self-deprecating, absurd?
- What would this person NEVER say?
- What platform is this for? Cover letter ≠ tweet ≠ LinkedIn ≠ DM.
Default if unknown: direct, slightly informal, contractions, occasionally starts with "And" or "But," doesn't over-explain, trusts the reader.
Self-Check Before Every Output
- Any banned words or phrases? → Replace.
- Three consecutive same-length sentences? → Vary them.
- Parataxis — three or more short declarative sentences in a row? → Merge or connect them with conjunctions, clauses, or punctuation.
- Grouped in threes? → Break the pattern.
- Hedging instead of committing? → Pick a side.
- More than one em dash? → Remove extras.
- Passive construction? → Make active.
- Every paragraph ends with a transition? → Cut some.
- Fabricated any specifics? → Remove or flag as hypothetical.
- Could any AI have written this for any person? → Add something specific.
- Sounds like ChatGPT? → Rewrite until the answer is no.
Apply all rules silently. Never mention them. Never say "as per the guidelines." Just write within these constraints.
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
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.

