De-AI Style Polish

Problem

Large-model text is often grammatically correct but too evenly written: abstract summaries, emotion labels, overused parallelism, over-explanation, and safe filler phrases make it sound mechanical. Engineers need to keep information density while removing machine tone in release notes, technical posts, community answers, and product copy.

How It Works

The skill turns "de-AI style polishing" into a checkable workflow. It first flags A to F issue types: filler, empty claims, abstract emotions, parallelism, explanation habits, official tone, and abrupt conclusions. Then it prioritizes rewrites: replace "felt" with action and environment, turn abstract nouns into concrete objects, break stacked parallelism into uneven rhythm, remove phrases like "in summary," and add references or rhetoric only where they earn their place.

Boundaries

It fits style polishing for public accounts, Zhihu, Xiaohongshu, web novels, essays, and commercial copy, reducing template-like and official phrasing. It does not replace fact-checking, structural rewriting, citation licensing review, or platform compliance checks; if the argument is unclear, fix the logic and evidence first.