This skill addresses the challenge undergraduate computer science students face in generating their first thesis draft, particularly in integrating project specifics, academic norms, and visual elements. It provides an automated workflow, leveraging AI assistants to transform fragmented materials into a structured .docx document, complete with diagrams and screenshots, significantly streamlining the initial writing phase.

How It Works

The lunwen-skill operates as an AI-driven assistant, designed to integrate with platforms like Codex and Claude Code. Its core methodology prioritizes "project facts first," analyzing code and documentation to establish a factual foundation. It then extracts stylistic and structural elements from university templates or sample papers to guide content generation. The process involves creating a target word count per chapter, writing content while adhering to these counts, and managing a reference pool with attention to language ratios. Novelty lies in its automated generation of diagrams (Mermaid/PlantUML) and system screenshots (via Playwright/Chrome CDP), culminating in a .docx draft that closes the loop on visual and textual requirements.

Quick Start & Requirements

Installation can be done via Codex using install-skill-from-github.py or by manually cloning the repository into the .codex/skills directory. Key prerequisites include Python 3.9+, an AI assistant environment (Codex native, Claude Code/Trae with compatibility layers), and potentially LibreOffice/Poppler for advanced document checks and Mermaid rendering capabilities for diagrams. Browser screenshot generation relies on additional environment support. Detailed installation instructions are available in ./INSTALL.md.

Highlighted Details

• Automated generation of thesis drafts driven by sample papers, opening reports, or university templates.
• Automatic inclusion and generation of flowcharts, ER diagrams, and project screenshots.
• Extraction of DOCX styles and structures for consistent formatting.
• Management of reference pools with checks on Chinese/English citation ratios.
• Integrated strategies for browser-based auto-screenshots and .docx output delivery.

Maintenance & Community

The repository does not explicitly detail specific contributors, sponsorships, or community channels like Discord or Slack. However, the comprehensive README and structured approach suggest active development.

Licensing & Compatibility

The project is released under the MIT License, which is permissive for commercial use and integration into closed-source projects. It offers compatibility layers for Claude Code and Trae, alongside native support for Codex.

Limitations & Caveats

Functionality requiring page-by-page rendering checks is dependent on the presence of LibreOffice or Poppler. Flowchart generation may be limited to source code or placeholders if Mermaid rendering capabilities are absent. Automated browser screenshots require additional environment setup.