STE80 Humanizer skill
A skill that edits AI-written English in two passes: it removes 26 AI tells and simplifies sentences after ASD-STE100
- 464 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill holds text editing rules, and the linter only reads files and prints findings.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/ste80-humanizerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill holds text editing rules, and the linter only reads files and prints findings.
This is third-party code. Review the repository files before installing.
What it does
STE80 Humanizer makes English text easy for people to read and keeps its meaning. The first pass removes 26 AI writing tells from Wikipedia's list in the blader/humanizer catalog: not-X-but-Y contrasts, chatbot phrases, run-ups, stacked hedges, fake depth. The second pass rebuilds sentences "80% of the way to ASD-STE100", the simplified English of aircraft maintenance manuals. One idea per sentence, active voice, instructions apart from descriptions. The method follows Andrej Karpathy's tips from 2 October 2026. There are three modes: normal editing, a light clean-up of tells only and strict STE for runbooks and procedures. The skill adds no facts, keeps numbers and degrees of certainty and flags contradictions in the source. A dependency-free linter `scripts/ste_check.py` is included.
Who it is for. Technical writers and developers who publish English docs, READMEs and posts with AI help and want to strip its typical feel.
Good fit when
- An AI draft of a README is full of long sentences and hype, and it should become dry and precise.
- An on-call runbook must read unambiguously, with one instruction per sentence.
- A LinkedIn post sounds like ChatGPT, while its numbers and wording must stay intact.
- You only want to know whether an email sounds like AI, without a rewrite.
Not a fit when
- The text is in Russian: a separate Russian-language editor fits better.
- You want text to slip past AI detectors: the authors do not aim for that.
- You need official ASD-STE100 compliance with the dictionary and trained writers.
Example request
Light touch: just remove the AI stuff from docs/intro.md.Limitations
STE targets explanatory text, so for fiction, speeches and brand copy the skill does only the first pass. The standard is English: other languages get the tells and structure rules, but not the English word lists. STE style is not STE compliance, which needs the official dictionary, trained writers and review. The repository contains neither the ASD-STE100 specification nor its dictionary, and the rules are paraphrased and simplified. The linter is a heuristic and not an AI detector.
How to disable. Remove the `ste80-humanizer@ste80-humanizer` plugin in Claude Code or the skill folder installed by `npx skills add`.
FAQ
How do I install STE80 Humanizer in Claude Code?
As a plugin: `/plugin marketplace add Dodaios/ste80-humanizer` and `/plugin install ste80-humanizer@ste80-humanizer`. For Codex, Cursor and OpenCode use `npx skills add Dodaios/ste80-humanizer` with the `--agent` flag.
How does STE80 Humanizer differ from blader/humanizer?
The first pass follows its catalog of 26 tells with new examples. The second pass adds simplified English rules: short sentences, active voice, separate instructions.
Will the skill help get past AI detectors?
No, the project has no such goal. The skill aims at text that people read easily.
How do I check text without an agent?
Run `python scripts/ste_check.py README.md`, adding `--strict` for strict mode. The linter flags AI tells, sentence length, semicolons, dashes and passive voice.
Can the edit match my own style?
Yes. Add a sample of your writing, and the skill matches its sentence length, words and punctuation.
Discussion
No questions or reviews yet.
Related
SkillsA skill that removes AI tone from work writing while keeping every fact, commitment and caveat at the same strength.
Humanizer (brandonwise)
humanizer
A skill and Node.js CLI that score English text for AI-writing patterns and statistics and suggest fixes on a 0 to 100 scale
An editor for AI-written Russian text: removes bureaucratese, cliches and 38 AI patterns while keeping facts intact
A skill that finds AI writing tells, rewrites the text in a chosen voice, and optionally scores it from 0 to 100
An editor skill for Chinese text: removes filler, repetition and template phrasing, keeping facts, certainty and author voice
Two skills for English prose: humanize rewrites a model draft in a natural voice, and ai-check flags machine-writing tells
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| STE80 Humanizer (this card) | no | 464 | |
| Humanizer | no | 68,355 | removes AI tells from Wikipedia's list; here the same list plus a second pass by simplified English rules |
| Humanizer RU | no | 210 | a Russian text editor; here rules for English modelled on ASD-STE100 |
README in short
The README shows a before-and-after example, tables of both passes with sample edits and three modes. It also reports four tests in the `claude plugin eval` format and lists sources.
README badge
Original SKILL.md
--- name: ste80-humanizer description: | Rewrite AI-sounding text so it reads like a clear, careful person wrote it, using the "80% of the way to ASD-STE100" method that Andrej Karpathy shared. Pass 1 removes the 26 AI tells from Wikipedia's "Signs of AI writing" (not-X-but-Y, one-line closers, forced triads, dashes everywhere, stock AI words, inflated claims, chatbot residue). Pass 2 rebuilds the text with the structure of Simplified Technical English: one idea per sentence, short sentences, active voice, simple tenses, plain words. Use it whenever the user wants text humanized, de-AI'd, made clearer, simpler or less robotic, asks for ASD-STE100 or STE, or wants docs, READMEs, emails, posts or answers tightened. license: MIT metadata: version: "4.0.0" author: "Dodaios" --- # STE80 Humanizer Rewrite AI-sounding text so it reads like a careful person wrote it, and so the reader gets it on the first pass. Keep what it says. Do not make anything up. The skill has two passes: - **Pass 1 (§1 to §26)** removes AI tells: habits that people recognize as machine writing. - **Pass 2 (S1 to S15)** rebuilds sentences with the structure of ASD-STE100, softened to 80%. Report your edits by these numbers. *Weak alone* means: act only when other tells are near it. ## The method Andrej Karpathy suggested asking an LLM to write "80% of the way to ASD-STE100". ASD-STE100 (Simplified Technical English, STE) is the controlled English of aircraft maintenance manuals. You already know the spec. Apply Issue 9 (2025). The full spec is too strict for normal prose. At 80%, keep all of the structure and only as much of the vocabulary as still sounds natural: | STE rule | 100% (strict) | 80% (default) | |---|---|---| | Sentence length | 20 / 25 words, hard limit | A few words over is fine if a split would break the meaning | | One idea per sentence | Always | Always | | Active voice, simple tenses | Always | Always | | Approved dictionary only | Yes | No. Use the plain common word |