humanizer-zh

Rewrites Chinese text to remove 24 documented AI writing patterns.

5|1|Updated Jul 31, 2026
One-click install
npx skills add https://github.com/shiker1996/wechat-editroom --skill humanizer-zh-shiker1996
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: humanizer-zh
Source: https://github.com/shiker1996/wechat-editroom/tree/main/skills/humanizer-zh
Command: npx skills add https://github.com/shiker1996/wechat-editroom --skill humanizer-zh-shiker1996

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI-generated Chinese text often carries detectable machine-writing traces such as exaggerated significance claims, vague attribution, formulaic triads, filler phrases, and overused connectives. This Skill identifies and rewrites those patterns so the text reads naturally while preserving meaning, tone, and Markdown heading structure. ## Core Features & Use Cases - 24-pattern detection: Scans text across four categories—content, language/grammar, style, and conversational filler—based on Wikipedia's Signs of AI writing guide. - Quality gate scoring: Rates rewrites on a 50-point rubric (directness, rhythm, trust, authenticity, concision) and rejects output below 35 points for rework. - WeChat article pipeline integration: Reads 03-draft.md and writes 04-humanized.md as Step 4 of the wechat-mp-topic-to-article workflow, with rules protecting hooks, conflict, and author voice in traffic-driven articles. - Use Case: Paste an AI-drafted WeChat article full of phrases like "此外" and "至关重要"; the Skill rewrites it with concrete details, varied sentence rhythm, and natural voice while keeping all headings intact. ## Quick Start Ask the assistant to humanize your pasted Chinese article text so it reads naturally and removes AI writing traces.

Frequently Asked Questions about humanizer-zh

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I remove AI writing traces from Chinese text?

Scan the text against the 24 documented AI writing patterns, then rewrite each flagged segment with concrete details and natural phrasing. Preserve the core meaning, match the intended tone, and keep all Markdown headings unchanged.

What AI writing patterns does this detection cover?

It covers 24 patterns in four categories: content patterns like exaggerated significance and vague attribution, language patterns like AI vocabulary and rule-of-three overuse, style patterns like em-dash and bold overuse, and conversational filler like sycophantic tone and generic positive conclusions.

Can this humanize WeChat official account articles?

Yes, it integrates as Step 4 of the wechat-mp-topic-to-article pipeline, reading 03-draft.md and writing 04-humanized.md. Special rules protect hooks, conflict, author stance, and punchy sentences in traffic-driven articles while only fixing AI-flavored phrasing.

Does rewriting AI text change the original meaning or facts?

No, the rewrite preserves core information and never adds new facts, quotes, numbers, sources, or calls to action. It only replaces AI-trace phrasing with natural equivalents and maintains the appropriate tone for the context.

What happens when the humanized text quality is too low?

Output scoring below 35 out of 50 is rejected and sent back for rewriting, with a note that AI traces remain too heavy. Scores of 35-44 must be revised to 45 or higher before delivery, and more than 5 AI vocabulary hits trigger a full rework.