Repolode
Submit
MrGeDiao

MrGeDiao/shuorenhua

A Chinese-first editing skill for Claude Code, Codex, Cursor and ChatGPT that strips AI-style filler from text while keeping facts and conditions intact.

View on GitHubWebsite
Stars
2,227
Forks
84
Open issues
1
Last commit
4 hours ago

What is it?

shuorenhua is an agent skill, a set of editing rules that runs inside tools such as Claude Code, Codex, Cursor and ChatGPT. You paste a draft and it returns a cleaner version: greetings with no content, announcements, empty thanks, and repeated sentences are removed, while numbers, dates, version numbers, conditions, commitments and the author's tone are kept.

It works in several modes. The default lets it delete, merge and reorder sentences. Bounded and in-place modes limit changes to wording within a sentence, which suits text where sentence structure matters. A separate plain-style request removes promotional wording when the user asks for it. Chinese is the primary target, and English input is supported as well. It is not an AI-authorship detector and makes no claim about passing AI detection.

The project ships with a benchmark of 123 cases, including cases where the input is already fine and should be left alone. Known limits are documented, such as conservative handling of evaluative praise and weaker results on long articles.

Key features

  • 01Removes filler greetings, announcements, empty thanks and repeated sentences.
  • 02Keeps numbers, dates, version numbers, conditions and commitments unchanged.
  • 03Offers structural, bounded and in-place modes to control how much the text can change.
  • 04Supports a review-only mode that marks problems without rewriting.
  • 05Installs with one npx command for Claude Code, Codex and Cursor, and also works as a ChatGPT GPT.

Use cases

  • Tidying a work status update before sending it to a team chat.
  • Trimming a Chinese draft email or announcement while keeping its dates and commitments.
  • Reviewing a text for filler without letting the tool rewrite it.
  • Cleaning up AI-generated Chinese copy so it reads more naturally.

Quick start

  1. 1

    Install the skill with npx (works for Claude Code, Codex, Cursor and other skills-enabled tools)

    npx skills add MrGeDiao/shuorenhua
  2. 2

    Paste your text into the chat and ask it to use 说人话 (shuorenhua) to remove filler

Things to know

  • Conservative by default: evaluative praise is kept unless you explicitly ask for it to be removed.
  • Long-form articles get fewer cleanups, and results vary across models.
  • One known test case showed a sentence with a delivery promise being wrongly deleted.

Topics

#agent-skill#chinese#writing#text-quality#prompt-engineering#rewriting#claude-code#codex#ai-writing#anti-slop

FAQ

What is shuorenhua?
A Chinese-first editing skill for Claude Code, Codex, Cursor and ChatGPT that strips AI-style filler from text while keeping facts and conditions intact. shuorenhua is an agent skill, a set of editing rules that runs inside tools such as Claude Code, Codex, Cursor and ChatGPT. You paste a draft and it returns a cleaner version: greetings with no content, announcements, empty thanks, and repeated sentences are removed, while numbers, dates, version numbers, conditions, commitments and the author's tone are kept.
Is shuorenhua free to use? What license does it have?
shuorenhua is open source under the MIT license. Check the license terms before using it in commercial products.
Is shuorenhua still maintained?
Actively maintained. The last commit to shuorenhua was on Oct 22, 2026.
What can I use shuorenhua for?
Tidying a work status update before sending it to a team chat. Trimming a Chinese draft email or announcement while keeping its dates and commitments. Reviewing a text for filler without letting the tool rewrite it. Cleaning up AI-generated Chinese copy so it reads more naturally.
How do I install shuorenhua?
Install the skill with npx (works for Claude Code, Codex, Cursor and other skills-enabled tools): npx skills add MrGeDiao/shuorenhua Paste your text into the chat and ask it to use 说人话 (shuorenhua) to remove filler
Where can I get shuorenhua?
The source code and installation instructions are on GitHub: https://github.com/MrGeDiao/shuorenhua

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects