Humanize Writing rewrites AI-generated text through three passes: it removes overused AI vocabulary, breaks common AI sentence structures, and adds more human texture. It also includes LinkedIn-specific rules and a 14-point quality checklist before it returns a rewrite.
Builders who want Claude Code, Cursor, Windsurf, or another LLM to rewrite text so it reads like a person wrote it.
You can turn obvious AI copy into text that sounds more direct, varied, and human without re-explaining the rules each time.
Pass 1 removes overused AI words, Pass 2 breaks common AI sentence patterns, and Pass 3 adds contractions, variation, and visible opinion.
A reference file with 36+ banned words, 10 structural patterns, tone tells, and content red flags.
Special guidance for rewriting LinkedIn posts so they do not read like generic AI content.
A 14-point checklist that runs before the rewrite is returned.
A one-command install that makes the skill available in Claude Code.
claude plugin marketplace add https://github.com/lguz/humanize-writing-skill
A prompt-based writing editor that rewrites AI-generated text to sound like a real human wrote it.
It catches the patterns that make AI writing obvious — the vocabulary, the sentence structures, the too-perfect tone — and fixes them through a systematic 3-pass editing process.
Works with any LLM: Claude, ChatGPT, Gemini, Cursor, Windsurf, or anything that accepts a system prompt. Also available as a one-command install for Claude Code users.
The system runs your text through three passes:
Pass 1: Kill the AI Vocabulary — Replaces words like "delve," "leverage," "tapestry," and 30+ other statistically overused AI words with simpler human alternatives.
Pass 2: Break the AI Structures — Eliminates structural patterns like parallel negation ("Not X, but Y"), tricolons (groups of three), em dash overuse, rhetorical Q+A, and mirror structures. These are stronger tells than vocabulary.
Pass 3: Add Human Texture — Varies sentence length, adds contractions, lets some thoughts stay unresolved, and makes the author's actual opinion visible.
Also includes special rules for LinkedIn posts and a 14-point quality checklist that runs before returning any rewrite.
The entire system is two markdown files. You can use them right now:
Copy the contents of both files into your LLM's system prompt or custom instructions, then paste any text you want humanized.
One command, and it works automatically whenever you ask to humanize text:
claude plugin marketplace add https://github.com/lguz/humanize-writing-skill
After that, just say things like "humanize this" or "make this sound less AI" and Claude Code will use the skill automatically.
Copy the skill files into your Claude Code skills directory:
mkdir -p ~/.claude/skills/humanize-writing/references
curl -o ~/.claude/skills/humanize-writing/SKILL.md https://raw.githubusercontent.com/lguz/humanize-writing-skill/main/skills/humanize-writing/SKILL.md
curl -o ~/.claude/skills/humanize-writing/references/ai-patterns-dictionary.md https://raw.githubusercontent.com/lguz/humanize-writing-skill/main/skills/humanize-writing/references/ai-patterns-dictionary.md
.cursor/rules/humanize-writing.md or .windsurfrules)Copy the contents of both markdown files into whatever "system prompt," "custom instructions," or "context" field your tool provides. The instructions are model-agnostic — they work with any LLM that can follow structured prompts.
skills/humanize-writing/
├── SKILL.md # 3-pass rewriting process + quality checklist
└── references/
└── ai-patterns-dictionary.md # Banned words, structural patterns, tone tells
Based on the principles behind Paul Graham's writing style: clarity over cleverness, directness over decoration. Good writing sounds like a smart person thinking out loud. The goal is invisible editing — the reader should never think about how something was written.
The AI patterns dictionary draws from:
MIT
Sign in to join the discussion.
No comments yet. Be the first to say what this is good for.
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.
Topic in, narrated explainer video out. A Claude Code / Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles and a chapter progress bar. Chinese or English; every frame drawn in code with Remotion.
Public repository for Agent Skills
Open-source AI job search: scan job portals, evaluate listings into a structured A-H report with a global 1-5 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
A skill to stop your coding agent from burying the answer. ADHD-friendly output.