Verified project record

op7418/Humanizer-zh

humanizer-zh is a Chinese-language Claude Code skill that edits empty, repetitive, and templated phrasing out of articles, comments, and docs while preserving facts, certainty levels, and the author's voice. It works from a checklist of 31 editing checkpoints and is activated with the /humanizer-zh command.

Project overview

It targets a concrete, recurring pain: AI-generated Chinese text full of formulaic, inflated, templated expressions, and its editing rules explicitly preserve negation, conditionals, attribution, scope, timing, and completion status, so 'possible' never becomes 'certain' and 'planned' never becomes 'done'.

Project type: Prompt Engineering
Use cases: Documents & Office
Deployment: Refer to project documentation
License: MIT

Best for

Creators and general users who write Chinese articles, comments, or docs and want AI-generated formulaic phrasing edited out while preserving facts and their own voice.
Users who already run Claude Code and can install skills.

Key capabilities

Edits empty talk, repetition, and templated expressions in existing Chinese articles, comments, and docs; sentences without issues may be left unchanged; output is a final rewritten draft.
31 editing checkpoints in A–F categories covering pseudo-comparison, forced triads, hedging stacks, authority-borrowing, formulaic formatting, chat leftovers, and Chinese-specific patterns such as long attributive chains and the common 'conduct plus verb' padding construction.
Preserves facts, negation, conditionals, attribution, scope, timing and certainty; can learn expression habits from author samples without importing their experiences, data, or opinions.
When asked to 'give suggestions, do not modify the file', the skill only reviews rather than rewriting.
After installation and reload, the skill activates when the user types /humanizer-zh in Claude Code.

Limitations and risks

Effectiveness depends on the executing agent/model reading and following the guideline; results may vary with the model used.
The skill needs Claude Code (or a compatible agent) that can read and execute the skill definition; it is not a standalone tool.

Getting started

Run npx skills add https://github.com/op7418/Humanizer-zh.git, then restart Claude Code or reload skills, then type /humanizer-zh in the conversation. Manual installation is copying a folder; overall setup difficulty is documented as easy.

Evidence and sources

README
README
README
README: npx skills add https://github.com/op7418/Humanizer-zh.git
README

Source updated at: 2026-09-25T05:25:36Z

Open original GitHub project