zouchenzhen/docx-template-translator-skill

A Codex skill for translating LaTeX, PDF, Markdown, or rough DOCX into specified Word template formats.

It is designed for scenarios such as graduation theses, theses, unit reports, and standard documents, especially suitable for tasks where pandoc’s default DOCX output is insufficient and where strict application of school or institution Word templates is required.

English: README.en.md

---

See the effect in one minute

The repository comes with a built-in minimum runtable example examples/minimal_markdown/, which runs in the repository root:

The script generates a sample template on site, running inspect / adaptive / finalize in sequence, and then Export PDF preview images in Windows + Word environment. The two images below are the same one copy sample.md and the same authentic Zhengzhou University thesis template ** generated, the first copy is for the main project adaptive pipeline, the second card only runs directly pandoc --reference-doc:

Project output: pandoc → adaptive pipeline → Word finalize

Direct pandoc baseline: only use pandoc --reference-doc

Both images use the same input and template, but the numbering context is different: the output for this project will be retained The actual template already contains prerequisite content and chapter numbers, so the sample section will follow the original chapter of the template (for example, displayed as 4.7 / 8.7 numbers, depending on the current content of the template); Direct pandoc baseline creates a document from sample.md and only reuses the style definition of --reference-doc. Therefore, numbering starts from Chapter 1 (e.g., 1.5).

Comparison shows: pandoc --reference-doc can reuse some styles, but it only knows the style definition, Not understanding the main text semantics and layout constraints in school templates; Charts/formulas/tables are easier to span pages and have loose spacing, The table is also not the three-line table effect required by the target template. This project will inspect the template first, then use it Adaptive pipeline maps the body, title, formulas, and tables, then exports them using Word finalize, Therefore, it is more suitable for strict school/unit templates.

The output image for this project is generated using the following command, then the main area is cropped from page 20:

When --template is not transmitted, run_example.py will be generated live using build_template.py. Lightweight sample template, with more simplified output but fully reproducible; --preview-pages Default "1-4", just on the demo template.