Remove AI Slop from Your Writing
Claude writes clean drafts fast. The problem is that clean AI drafts all sound alike. Readers spot "It's not X. It's Y." and "Here's what nobody tells you" in one line and scroll past.
No AI Slop is a free, MIT-licensed Skill by Peter Yang. It is a single editing ruleset: one SKILL.md with the rules and one eval.md with checks the Skill runs on its own output. It needs no API keys, scripts, or connectors.
What it does
The Skill has two modes.
| Mode | How to trigger it | What you get back |
|---|---|---|
| Edit (default) | /no-ai-slop + your draft | The edited draft plus a What changed list |
| Detect | /no-ai-slop is this slop? + your draft | Each pattern it found, the quoted line, and a short fix. No rewrite, no score, no guess about whether AI wrote it |
It makes the smallest edit that fixes the draft. Your vocabulary, humor, bluntness, and honest uncertainty stay. It cuts:
- Binary contrasts: "It's not X. It's Y."
- Throat-clearing openers: "Here's the thing," "Let me be clear"
- Faux-insight setups: "What nobody tells you," "The part everyone misses"
- Colon reveals: "The best part: it learns."
- Superficial analysis: trailing "highlighting," "underscoring," "showcasing"
- Importance puffery: "a testament to," "marks a pivotal moment"
- Weasel attribution: "experts agree," "studies show"
- Synonym cycling, negative listing, dramatic fragments, fake-profound endings, recap endings
- Formatting slop: emoji headers, random bold, bullets where two sentences would do, em dash clusters
It also bans a short word list (delve, leverage, robust, game changer, supercharge, harness, and others) and trims empty adverbs and filler phrases unless they carry your voice.
Install
Claude Code, Codex, or any agent that can install Skills:
npx skills add petergyang/no-ai-slop --skill no-ai-slop --global --yes
Or paste this into your coding agent:
Install the /no-ai-slop skill globally from https://github.com/petergyang/no-ai-slop
It is also listed as a plugin in ChatGPT.
Use case 1: Final pass on every LinkedIn post
Write the post however you normally do, with Claude, a writing Skill, or by hand. Then run the cleanup as a separate step.
/no-ai-slop This is a LinkedIn post for B2B founders. Keep the first line as written.
(paste post)
Why a separate step: writing Skills built on viral formulas produce the exact patterns this Skill removes. Drafting and cleaning in one prompt gives you a half-clean draft. Two passes give you the formula's structure without its tells.
Protect your hook. A contrarian first line can look like a binary contrast. If the hook is working, tell the Skill to leave it alone, as in the prompt above.
Use case 2: Audit a draft before you touch it
Use detect mode on a teammate's draft, a ghostwritten post, or a freelancer's article.
/no-ai-slop is this slop?
(paste draft)
You get a list of quoted lines with fixes. The writer can check each one. This works better than an AI detector score because every finding points to a line you can change.
Use case 3: Study your swipe file
Run detect mode on a post that performed well and one that flopped. Compare the lists.
/no-ai-slop is this slop? Post A:
(paste)
Post B:
(paste)
You will often find that the post that flopped has more of these patterns. Treat that as a hypothesis to test on your own posts, not a rule.
Use case 4: Clean long-form without losing your voice
Newsletters, blog posts, and YouTube scripts drift into recap endings and "In conclusion" paragraphs. Give the Skill the audience and the goal so it knows what to keep.
/no-ai-slop Newsletter for agency owners. Goal: they book a call.
Keep my asides and the story in paragraph 2.
(paste draft)
Read the What changed list. If it reorganized anything, it explains why. Undo any change that made the piece sound less like you.
Use case 5: Sales copy and cold email
Landing pages and outreach emails fill up with puffery ("a game changer for your team") and weasel claims ("studies show"). The Skill replaces those with specifics from your draft or asks you for the missing fact. It does not invent stats or sources.
/no-ai-slop Cold email to a marketing director. Under 120 words.
(paste email)
If it asks for a number or a customer name, supply a real one or cut the claim.
Use case 6: Chain it into your own Skills
If you maintain writing Skills, add a last step that hands the draft to this one.
## Final step
After the draft is approved, run /no-ai-slop on it in edit mode.
Keep the hook line unchanged. Show the What changed list.
Keep your formula Skill focused on structure and let this Skill handle cleanup. Each Skill stays small and easy to debug.
When not to use it
- Satire or deliberate hype. The Skill can also write slop on purpose ("Draft an AI slop post about [topic]"). Use that for jokes, not client work.
- Short DMs with a strict format. A three-line DM has little room for slop. A message-specific editor fits better.
- Legal, medical, or compliance copy. The Skill edits for readability. Wording in those documents may need to stay exactly as approved.
Start with one small test
- Pick one post you already published.
- Run detect mode and read every finding.
- Run edit mode on the same post.
- Read both versions aloud. Keep the edit only if it still sounds like you.
Built by Joon at AI Topia. joon@getaitopia.io YouTube @joonahn_ai · LinkedIn
See also Useful External Skills.
Want this set up for your business?
We turn these workflows into working marketing and sales systems.