Strip the tells of AI writing

Make AI writing sound human is an Agent Skill: a written method your AI applies on demand, in Zealos or in any client that reads SKILL.md.

AI writing gives itself away structurally before it does lexically: paragraphs of identical length, lists that always contain three items, and a closing summary that restates a body the reader just finished.

What is in this skill

nameno-slop
descriptionStrip the statistical fingerprints of AI writing from a draft. Use as a final pass over any prose before it ships, and whenever text reads as generic, padded or oddly cheerful.
allowed-toolsnone — this skill adds judgement, not reach
compatibilityZealos, Claude Code, Cursor, Codex and any client that reads SKILL.md
licenseFree to use and edit

How it works

  1. 1Break the rule of three. Real writing has lists of two and four; a page where every list has exactly three items is a page a model wrote.
  2. 2Vary paragraph length deliberately, and allow one-sentence paragraphs — they are usually the strongest line on the page.
  3. 3Delete the participial tails. Any clause ending in ensuring seamless collaboration across teams adds nothing and appears everywhere.
  4. 4Cut twenty percent and see what breaks. If nothing does, the cut was correct and the draft was padded.

The skill itself

---
name: no-slop
description: Strip the statistical fingerprints of AI writing from a draft. Use as a final pass over any prose before it ships, and whenever text reads as generic, padded or oddly cheerful.
---

# De-slop the draft

Run this last, over finished prose. It removes the tells that make writing read
as machine-generated even when the content is right.

## Structural tells

- **Rule of three everywhere.** "Fast, reliable, and secure." Real writing has
  lists of two and four. Break the pattern.
- **Every paragraph the same length.** Vary it. One-sentence paragraphs are
  allowed and are often the strongest.
- **Symmetric sections.** If each heading has exactly three bullets, a template
  is showing.
- **Summary that restates the body.** Cut it. The reader was there.

## Sentence tells

- **"It's not just X — it's Y."** Almost always deletable.
- **"Whether you're X or Y."** Pick one reader.
- **Hedging stacks.** "It may potentially help to consider" → "Try".
- **Participial tails.** "…, ensuring seamless collaboration across teams."
  These add nothing and appear everywhere.
- **Em-dash pairs used as a verbal tic.** Keep the ones that earn their place.
- **"Delve", "tapestry", "testament", "underscore", "pivotal", "realm",
  "landscape", "navigate the complexities".**

## Tone tells

- Unearned enthusiasm. No exclamation marks unless something is genuinely good news.
- Flattering the reader ("Great question!").
- Apologising for length, then not shortening.

## Test

Cut 20% and see what breaks. If nothing does, the cut was correct — keep it.

Who uses it

Marketer

Copy stops reading as generated without anyone having to rewrite it by hand, which is the difference between shipping AI drafts and quietly abandoning them.

Founder

The LinkedIn post you generated stops sounding like every other LinkedIn post generated that week.

Account executive

Prospect-facing emails lose the tics that make a recipient assume nobody read what they wrote before replying.

Frequently asked questions

What are the tells of AI writing?

Symmetric structure, the rule of three everywhere, hedging stacks, it is not just X it is Y, participial tails, unearned enthusiasm, and words like delve, tapestry, underscore and seamless.

Why is this a separate skill from the voice skill?

Because they run at different times. The voice skill shapes the draft; de-slopping is a final pass over finished prose, including prose a human wrote.

Does it work on long documents?

Yes, and structural tells are more visible there — a twelve-section document where every section has three bullets is the clearest signal of all.

Related skills

Triage your inbox by deal impact

AI email triage is worth having only when it sorts by consequence rather than by recency: what blocks a deal, what someone is waiting on you for, and what can be read at the end of the week.

EmailGmailAccount executives/email-rules

Write a follow up that gets a reply

A follow-up gets a reply when it opens with the specific commitment made on the call, stays under 120 words, and asks for exactly one thing the recipient can answer in a sentence.

EmailGmailMeetings/email-rules

Make AI writing sound like your company

A writing-voice skill is the highest-leverage thing you can give an AI, because it changes every sentence the product produces and it needs no tools, no integrations and no data.

WritingEmailProposals/our-writing-voice

Run your after call workflow automatically

After-call work is the twenty minutes nobody books: the CRM update, the recap, the follow-up, the task for the thing you promised — and it is the first work to disappear on a busy day.

MeetingsCRMEmail/call-recap-format

Write a cold outreach sequence

A cold sequence works when each touch carries new information: the second email should not exist unless it says something the first one did not, and most sequences fail this test by touch three.

ProspectingEmailSDR/email-rules

Write a linkedin connection note

A connection note has about three hundred characters and one job: prove you know who they are, without the sentence that reveals you found them in a filtered list.

ProspectingSDRWriting/our-writing-voice

Skills, in your workspace

Write it once. Every agent follows it.

Twelve built-in skills ship with every workspace, and you can draft your own with AI or import any public SKILL.md from GitHub.

7-day free trial. Set up in minutes. Cancel anytime.