Humanizer Pro

A Claude Code / Claude Agent skill that transforms AI-generated text into authentic human writing. It detects and strips robotic patterns, swaps out overused AI vocabulary, checks statistical tells, and injects real personality.

What it does

• Detects 24 AI writing patterns (significance inflation, promotional language, filler phrases, sycophantic tone, etc.)
• Flags 500+ AI vocabulary terms across three severity tiers
• Analyzes statistical signals (burstiness, type-token ratio, sentence-length variance)
• Removes chatbot artifacts and elegant-variation habits
• Injects personality through parenthetical asides, strategic typos, tangents, and asides

Works on social posts, articles, emails, marketing copy, and documentation — anything that needs to sound like a person wrote it.

Pattern set is based on Wikipedia's "Signs of AI writing", Copyleaks research, and real-world analysis.

Install as a Claude skill

Copy this folder into your Claude skills directory:

git clone https://github.com/Kevjade/humanizer-pro.git
cp -r humanizer-pro ~/.claude/skills/humanizer

Then invoke it in Claude Code by asking to "humanize" text, or with /humanizer.

Use the analyzer script directly

The statistical analyzer runs standalone (no dependencies, Node.js):

node scripts/humanize.js "your text here"
# or pipe a file
node scripts/humanize.js < draft.md

It reports an AI-likelihood score plus a breakdown of burstiness, type-token ratio, and sentence uniformity.

Repo layout

SKILL.md                          # skill instructions + quick start
references/
  patterns.md                     # the 24 AI patterns
  vocabulary.md                   # 500+ flagged terms by tier
  statistical-signals.md          # burstiness, TTR, sentence variance
  personality-injection.md        # asides, typos, tangents
  examples.md                     # before/after rewrites
scripts/
  humanize.js                     # standalone statistical analyzer

License

MIT — see LICENSE.