Repo of the Day
Dodaios/ste80-humanizer
Published: Nov 16, 2026
Open repository ↗STE80 Humanizer: agent skill that removes AI tells and rewrites text 80% of the way to ASD-STE100 (Simplified Technical English), the method Andrej Karpathy shared
Summary
STE80 Humanizer is an MIT-licensed agent skill that rewrites text in two passes: it removes 26 common AI-writing tells, then rebuilds sentences toward ASD-STE100. It provides default 80% STE, light-touch, and strict modes, plus a dependency-free Python checker for measurable style issues.
What it is useful for
The skill is aimed at technical writing that needs to be easier to read without losing facts. Its default mode removes phrases such as “Great question!” and “not just X, but Y,” then shortens sentences, uses active voice, and keeps instructions separate from descriptions. The README’s Kubernetes example becomes: “A liveness probe tells the kubelet if a container is healthy. If the probe fails, the kubelet restarts the container.”
Use it for READMEs, introduction docs, runbooks, technical explanations, and emails. Choose light mode when you only want the worst AI phrasing removed. Choose strict mode for procedures and safety notes. You can supply a sample of your own writing so the rewrite matches your sentence length, vocabulary, and punctuation.
This is a readability tool, not a way to bypass AI detectors. The skill reports edits by rule number, such as “§1, §8, S2,” which makes review against the documented catalogs possible.
How engineers can use it
Install it in the current project with the documented skills CLI command:
npx skills add Dodaios/ste80-humanizer
For a global install for Claude Code, use:
npx skills add Dodaios/ste80-humanizer --global --agent claude-code
The README also documents global installs for Codex, Cursor, and all supported agents, plus Claude Code plugin commands:
/plugin marketplace add Dodaios/ste80-humanizer
/plugin install ste80-humanizer@ste80-humanizer
After installation, run /ste80-humanizer and paste the draft, or make a request such as “Rewrite this README 80% of the way to ASD-STE100” or “Light touch: just remove the AI stuff from docs/intro.md.”
To review a file with the bundled checker:
python scripts/ste_check.py README.md
python scripts/ste_check.py --strict docs/runbook.md
cat draft.txt | python scripts/ste_check.py -
The checker flags chatbot phrases, stacked hedges, passive voice, long sentences, dashes, and semicolons. It is heuristic, not an AI detector or compliance checker.
Important limits: STE mode applies to explanatory text; fiction, speeches, and brand copy receive only Pass 1. Other languages use the tell and structure rules, but not the English word lists. The repository paraphrases and simplifies the STE rules and does not include the official specification or dictionary. “STE style” is not official ASD-STE100 compliance, which requires the official dictionary, trained writers, and review.