An agent skill that audits and rewrites prose to remove common AI writing patterns, with detect-only and edit-in-place modes.
larashero3-dotcom/lieflat-gongwen
An agent skill distilled from a 1.02-million-character corpus of Chinese official documents, turning official-writing style into measurable parameters for AI agents.
- Stars
- 1,536
- Forks
- 211
- Open issues
- 0
- Last commit
- 2 months ago
What is it?
This is a writing skill for AI agents such as Codex, Claude Code and Moxt, aimed at drafting Chinese government and organizational official documents. It covers seven styles in two families: research reports, leadership speeches, work opinions, experience reports, work plans and experience summaries in the official-document family, plus a shorter party-building experience piece.
Instead of relying on fixed templates, the skill derives its guidance from statistics over a large corpus of real documents, such as sentence length, enumeration density, heading structure and paragraph patterns. The repository includes a parameter card, heading and sentence reference libraries, and a Python self-check script that flags structural mismatches against the target genre, using only the standard library.
The skill excludes statutory formats with fixed structures such as notices, requests, replies and letters. The source corpus is not distributed due to copyright, so the included synthetic sample essays serve as structural references. Content accuracy and political wording still require human review.
Key features
- 01Covers seven official-document styles across two genre families.
- 02Provides a parameter card with per-genre statistical ranges.
- 03Includes a Python self-check script that flags genre-structure mismatches.
- 04Offers heading and sentence reference libraries for each document type.
- 05Works as an agents.md-compatible skill for Codex, Claude Code and Moxt.
Use cases
- Drafting a work plan or research report that should match real official-document conventions.
- Checking whether an AI-generated official document has the right heading and sentence structure for its genre.
- Teaching writing staff the stylistic differences between official-document types.
Quick start
- 1
Clone the repository into your agent's skills directory
git clone <this-repo> ~/.claude/skills/lieflat-gongwen - 2
Load the skill in a compatible agent (Codex, Claude Code, or Moxt) and follow the six-step workflow in SKILL.md
Things to know
- Licensed under PolyForm Noncommercial 1.0.0, so commercial use requires separate permission from the author.
- The source corpus is not distributed due to copyright, so the frequency figures cannot be independently verified.
- Some document types rely on small samples (10 work plans and 8 experience summaries), so their parameters are only a rough guide.
Topics
FAQ
- What is lieflat-gongwen?
- An agent skill distilled from a 1.02-million-character corpus of Chinese official documents, turning official-writing style into measurable parameters for AI agents. This is a writing skill for AI agents such as Codex, Claude Code and Moxt, aimed at drafting Chinese government and organizational official documents. It covers seven styles in two families: research reports, leadership speeches, work opinions, experience reports, work plans and experience summaries in the official-document family, plus a shorter party-building experience piece.
- Is lieflat-gongwen free to use? What license does it have?
- lieflat-gongwen is open source under the Other license. Check the license terms before using it in commercial products.
- Is lieflat-gongwen still maintained?
- Updated within 6 months. The last commit to lieflat-gongwen was on Sep 5, 2026.
- What can I use lieflat-gongwen for?
- Drafting a work plan or research report that should match real official-document conventions. Checking whether an AI-generated official document has the right heading and sentence structure for its genre. Teaching writing staff the stylistic differences between official-document types.
- How do I install lieflat-gongwen?
- Clone the repository into your agent's skills directory: git clone <this-repo> ~/.claude/skills/lieflat-gongwen Load the skill in a compatible agent (Codex, Claude Code, or Moxt) and follow the six-step workflow in SKILL.md
- Where can I get lieflat-gongwen?
- The source code and installation instructions are on GitHub: https://github.com/larashero3-dotcom/lieflat-gongwen
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A community-curated open-source library of prompts for AI chat models, with a website, CSV/Hugging Face datasets, CLI and MCP server integration.
A locally installed, power-user front end for chatting with many LLM APIs, image generators and TTS voice models.
An open format and specification for packaging instructions, scripts and resources into SKILL.md folders that AI agents load on demand.
A curated collection of 161+ ready-to-use Claude Code subagent definitions across 10 categories for development, infrastructure, security, data and more.
A compact, human-readable encoding of JSON data designed to reduce token usage when passing structured data to LLM prompts.