What problem does it solve?
Chinese text generated by AI often reads as stiff, formulaic, and full of clichés like "在……的今天" or "综上所述". This Skill quantifies those AI writing traces with an explainable 0-100 score, then guides targeted rewriting so the text reads like a real person wrote it, for WeChat official accounts (gzh) or Xiaohongshu (xhs).
Core Features & Use Cases
- Measurable AI-trace scoring: A deterministic Python script scores text across six signals (connector density, cliché phrases, concreteness deficit, sentence burstiness, list/dash overuse, paragraph uniformity) and returns the exact offending spans to fix.
- Score-rewrite-recheck loop: Rewrite against the highest-weighted signals, re-score, and iterate up to 3 rounds until the score drops to 55 or below.
- Platform-specific tones: Separate weight profiles and rewriting playbooks for gzh (restrained, coherent long-form) and xhs (short, punchy, emoji-friendly fragments), with before/after examples in the reference playbook.
- Use Case: After drafting a WeChat article that feels robotic, run the scorer to find it hits 78 with heavy cliché and connector signals, then rewrite using concrete numbers, first-person voice, and varied sentence lengths until the verdict becomes human-like.
Quick Start
Score this Chinese article for AI writing traces and rewrite it in a natural gzh tone until the score is 55 or below.
