hyacehila/humanizer-zh-next

Humanizer-zh-next is a skill for AI coding assistants that rewrites Chinese text to sound less like it was written by an AI and more like natural human writing. It plugs into agent tools such as Claude that support a skills system, and once installed you can ask it to edit or review a piece of Chinese text to remove the telltale patterns of AI generated writing while keeping the original facts and tone. The project builds on an earlier Chinese version and absorbs rule structure and detection ideas from an English language sibling project, along with some structural checks from a similar tool and guidance drawn from Wikipedia's writing about signs of AI text. It focuses on patterns common in Chinese AI writing: trendy business jargon, guided tour phrases like introductions that promise to explore a topic in depth, false authority phrases, vague upbeat conclusions about the future, and the kind of tidy before and after narration often seen in technical documentation. Rather than treating formal, academic, or clean grammar as a red flag by itself, the skill targets writing that feels hollow, inflated, overly templated, or unlike the author's usual voice. Its stated method is to rewrite passages rather than delete them, keeping the same number of paragraphs, and to follow a workflow of drafting, checking what makes text read as AI generated, then producing a final version. It also applies guardrails so that personality is only added to blog posts and opinion pieces, while technical and reference text stays neutral. Installation can be done by asking a supported AI coding assistant to install it directly from the repository URL, through a one line command using a skills installer tool, by cloning the repository into the assistant's skills folder, or by manually copying the files so the main skill file sits at the root of a folder named humanizer-zh-next. The README also lists two related projects for editing English text and academic writing, noting this project intentionally stays focused on general Chinese writing rather than academic use. It is released under the MIT license.