llms.txt git:20260714.e2b94eeA

Immutable. This exact content is served forever at /api/v1/blob/961fa4ee35555535.

# Humanizer — AI Writing Pattern Detector and Rewriter

> Humanizer is a pure-Markdown Claude Code skill that detects 53 AI writing patterns and rewrites text with human rhythm, voice, and burstiness. 5 named voice profiles. 0-100 AI-tell score. Zero-dependency Markdown core plus an optional metrics CLI and CI gate. No network calls, no binaries. Install in 5 seconds with a single curl command.

## Overview

- [README](https://github.com/Aboudjem/humanizer-skill/blob/main/README.md): Full documentation — install, usage, voice profiles, all 53 patterns, science, comparisons
- [SKILL.md](https://github.com/Aboudjem/humanizer-skill/blob/main/skills/humanizer/SKILL.md): The skill source — the only file needed to run humanizer
- [AGENTS.md](https://github.com/Aboudjem/humanizer-skill/blob/main/AGENTS.md): Harness context for Claude Code, Cursor, Copilot, Codex CLI, and 20+ other AI editors
- [Examples](https://github.com/Aboudjem/humanizer-skill/tree/main/examples): Before/after humanization examples with pattern annotations

## Install

- [Project-scoped install](https://github.com/Aboudjem/humanizer-skill#get-started): `mkdir -p .claude/skills/humanizer && curl -sL https://raw.githubusercontent.com/Aboudjem/humanizer-skill/main/skills/humanizer/SKILL.md -o .claude/skills/humanizer/SKILL.md`
- [Global install](https://github.com/Aboudjem/humanizer-skill#get-started): `mkdir -p ~/.claude/skills/humanizer && curl -sL https://raw.githubusercontent.com/Aboudjem/humanizer-skill/main/skills/humanizer/SKILL.md -o ~/.claude/skills/humanizer/SKILL.md`
- [Plugin marketplace](https://github.com/Aboudjem/humanizer-skill#get-started): `claude plugin marketplace add Aboudjem/humanizer-skill`

## Usage

- [Basic rewrite](https://github.com/Aboudjem/humanizer-skill#usage): `/humanizer "your text"` — rewrite with default voice
- [Detect mode](https://github.com/Aboudjem/humanizer-skill#usage): `/humanizer "text" --mode detect --score` — scan only, get 0-100 AI-tell score
- [Voice profiles](https://github.com/Aboudjem/humanizer-skill#voice-profiles): `--voice casual|professional|technical|warm|blunt`
- [File editing](https://github.com/Aboudjem/humanizer-skill#usage): `/humanizer --file docs/README.md --voice technical` — edit a file in place
- [Aggressive mode](https://github.com/Aboudjem/humanizer-skill#usage): `/humanizer "text" --aggressive --iterate 3` — convergence mode

## The 53 AI Patterns

- [Content Patterns P1–P8](https://github.com/Aboudjem/humanizer-skill#all-53-patterns): Significance inflation, promotional language, AI vocabulary ("delve", "leverage")
- [Language and Style P9–P18](https://github.com/Aboudjem/humanizer-skill#all-53-patterns): Negative parallelisms, em dash overuse, structured list syndrome
- [Communication P19–P21](https://github.com/Aboudjem/humanizer-skill#all-53-patterns): Chatbot artifacts, sycophantic tone
- [Filler and Hedging P22–P30](https://github.com/Aboudjem/humanizer-skill#all-53-patterns): Filler phrases, uniform sentence length, comprehensive overview openers
- [Emerging Patterns P31–P43](https://github.com/Aboudjem/humanizer-skill#all-53-patterns): Treadmill effect, infomercial hooks, elegant variation
- [Craft and Forensic P44–P53](https://github.com/Aboudjem/humanizer-skill#all-53-patterns): False agency, diff-anchored writing, reasoning-chain artifacts, unicode obfuscation, hedged-enumeration openers

## Science and Research

- [Burstiness](https://github.com/Aboudjem/humanizer-skill#the-science): Sentence length variation — humans vary wildly, AI averages 18 words. Detectors measure this.
- [Perplexity](https://github.com/Aboudjem/humanizer-skill#the-science): Word predictability — AI picks the statistically likely next word. Humans don't.
- [RAID Benchmark](https://github.com/Aboudjem/humanizer-skill#the-science): Structural paraphrasing drops DetectGPT accuracy from 70.3% to 4.6% (ACL 2024)
- [Research sources](https://github.com/Aboudjem/humanizer-skill#the-science): GPTZero, NeurIPS 2023, SSRN stylometric studies, Washington Post 328K message analysis

## Comparisons

- [vs QuillBot](https://github.com/Aboudjem/humanizer-skill#vs-alternatives): QuillBot swaps words; Humanizer fixes structure, rhythm, and burstiness
- [vs Undetectable.ai](https://github.com/Aboudjem/humanizer-skill#vs-alternatives): Paid, closed, no pattern detection, no voice profiles
- [Open source advantage](https://github.com/Aboudjem/humanizer-skill#vs-alternatives): 53 patterns vs 0, 5 voices vs 3, free vs $10–20/mo, explains every change

## Optional

- [Metrics CLI](https://github.com/Aboudjem/humanizer-skill/blob/main/cli/README.md): Zero-dependency Node tool that computes burstiness, TTR, trigram repetition, AI-vocab density, and a 0-100 score, with a CI quality-gate and GitHub Action
- [Chinese README](https://github.com/Aboudjem/humanizer-skill/blob/main/README.zh-CN.md): Simplified-Chinese docs positioned for humanizing English writing
- [CHANGELOG](https://github.com/Aboudjem/humanizer-skill/blob/main/CHANGELOG.md): Version history
- [CONTRIBUTING](https://github.com/Aboudjem/humanizer-skill/blob/main/CONTRIBUTING.md): How to add new patterns
- [Landing page](https://humanizer-skill.vercel.app): Interactive demo in the browser
- [Plugin manifest](https://github.com/Aboudjem/humanizer-skill/blob/main/.claude-plugin/plugin.json): Claude Code plugin metadata