Humanizer (Chinese) Skill

git clone https://github.com/op7418/Humanizer-zh ~/.claude/skills/Humanizer-zh

humanizer-zh solves a real problem: Chinese AI text has very specific tells. The “首先 / 其次 / 综上所述” structure. The over-explained transitions. The lack of mid-sentence reversal. The absence of personal hesitation.

This skill rewrites without losing meaning.

What it adjusts

• Connective overuse — strips “首先 / 其次 / 综上 / 总而言之” patterns
• Sentence rhythm — breaks long uniform sentences, adds variation
• Lexical entropy — replaces clichés with less predictable phrasing
• Voice signals — adds hesitation, side comments, mid-sentence pivots
• Formatting tells — softens overly tidy bullet structure

Where it shines

• Repurposing AI drafts for WeChat / public-facing channels
• Translation cleanup — translated Chinese often reads “AI-ish” too
• Bypass detection — for legitimate use (e.g., your own AI-assisted draft that you want to read more naturally)
• Voice consistency — match an existing author’s flow

Why it matters

In 2026, Chinese readers can spot AI text in three sentences. Audiences discount AI-flavored writing on credibility, viral potential, and emotional weight. Humanizing isn’t about deception — it’s about restoring the voice that AI flatness erases.

Composes with

• wechat-article-writer — pipe the output of any drafted article
• doc-to-markdown — clean extracted document text before re-publishing

→ Install on GitHub

Use humanizer-zh on this AI-generated Chinese article. Break up the
"首先...其次...再次...最后..." structure. Vary sentence length. Add
some hesitation, contradiction, or hands-on detail. Don't change facts.