Humanizer: Open Source Claude Code Skill to Humanize AI Text
Humanizer – A Simple Open‑Source Claude Code Skill for Human‑Sized Text
If you’ve ever read a paragraph that felt a little too polished or artificial, you’re not alone. Large language models (LLMs) are great at generating fluent sentences, but their output often contains tell‑tale signs that the text was machine‑written. Humanizer is a lightweight, open‑source skill for Claude Code that scans user‑supplied prose and eliminates 24 common AI‑writing patterns identified by the WikiProject AI Cleanup team.
Why it matters – Search engines and human readers increasingly value authenticity. By passing Humanizer’s filters, your content is less likely to be flagged by plagiarism or style checks and feels warmer to the eye.
1. What exactly does Humanizer do?
Humanizer applies a rule‑based checklist to incoming text and delivers a cleaned version. The pattern list is split into three categories:
| Category | Pattern # | Example of AI‑style | Humanized |
|---|---|---|---|
| Content | 1 | “marking a pivotal moment in the evolution of… ” | “was established in 1989 to collect regional statistics” |
| Language | 7 | “Additionally… testament… landscape” | “also… remain common” |
| Style | 13 | “institutions—not the people—yet this continues—” | “institutions, not the people, yet this continues.” |
The skill ships with a concise documentation block covering all 24 patterns, and the README contains a full before/after example that demonstrates the transformation.
2. How to install
a. Clone the repository directly to your Claude Code skills folder
mkdir -p ~/.claude/skills
git clone https://github.com/blader/humanizer.git ~/.claude/skills/humanizer
b. Manual install / update
If you already have the repo or just the SKILL.md file, copy it into a new folder:
mkdir -p ~/.claude/skills/humanizer
cp SKILL.md ~/.claude/skills/humanizer/
No additional dependencies are required – it’s a single SKILL.md file that Claude Code parses.
3. Using the skill in Claude Code
Once installed, you can invoke Humanizer in any Claude Code conversation with:
/humanizer
[Paste your text here]
Alternatively, add a natural‑language prompt:
Please humanize this text: “The future looks bright… ”
The output will be a new paragraph free of the 24 flagged patterns.
4. Real‑world examples
Before (AI‑sounding)
“AI coding assistants speed up some tasks. In a 2024 study by Google, developers using Codex completed simple functions 55% faster than a control group…”
After (Humanized)
"The productivity claims are hard to verify. The 2024 Uplevel study found no statistically significant difference in pull‑request throughput between teams with and without AI assistants."
The example demonstrates how Humanizer removes overly optimistic phrasing, unnecessary hedging, and promotional buzz‑words. The transformation leaves the core facts intact while making the tone genuine.
5. How it works under the hood
The SKILL.md file contains a series of #patterns directives. Each directive lists a regular‑expression matching AI‑style text and a replacement snippet. The skill runs as a pre‑processor inside Claude Code, scanning the user’s block of text and applying patterns in the order they appear.
Because the logic is declarative, contributors can easily add new rules. The community‑maintained list is updated regularly – the last commit in January 2026 added a new regular expression for curly‑quote replacement.
6. Community and impact
With 2.4k stars, 190 forks, and a modest watch‑list, Humanizer shows that there is genuine demand for lightweight tools that clean up AI output. The project is released under the MIT license, making it free to incorporate into commercial products or personal workflows.
Feel free to:
- Fork and extend the rule set
- Contribute to the README with additional before/after examples
- Report edge‑cases on GitHub or via the skill’s issue tracker
7. Takeaway
Humanizer proves that a small, well‑designed skill can significantly improve the quality of AI‑generated text. By incorporating the 24‑pattern checklist into your writing process, you:
- Increases reader trust – the prose feels human.
- Reduces compliance risk – avoids repetitive or promotional phrasing that can trigger content‑policy violations.
- Provides an easy, repeatable pipeline – no need for manual editing or costly APIs.
Whether you’re a freelance editor, a content‑marketing team, or a developer building tools that generate natural‑language output, Humanizer is a must‑try addition to your toolbox.
Want to give it a try?
Install, test the skill, and let us know how it transformed your next AI‑generated draft.