Codex Skill to generate Word documents based on your brand templates
English summary
The post describes a task to repeatedly generate Office documents (DOCX, PPTX, XLSX) using OpenAI's Codex while strictly adhering to pre-approved company brand templates. Initial tests with Claude's document-generation skills proved unreliable for preserving all template elements. The author built a custom solution over three days, leveraging AI to extract template characteristics and reuse them faithfully. The solution is now open-sourced on GitHub. The key insight is that AI needs a robust process to extract and reuse template features when generating variable-content documents.
Chinese summary
该帖子描述了一项任务：使用OpenAI的Codex重复生成Office文档（DOCX、PPTX、XLSX），同时严格遵循公司预批准的品牌模板。最初测试Claude的文档生成技能时，发现其在保留所有模板元素方面不可靠。作者用了三天时间构建了一个自定义解决方案，利用AI提取模板特征并忠实地复用它们。该解决方案现已开源在GitHub上。关键见解是，AI需要一种稳健的过程来提取和复用模板特征，以生成内容可变的文档。
Key points
Task required preserving every pre-approved design element, layout, style, and image from company templates.
任务需要保留公司模板中所有预先批准的设计元素、布局、样式和图像。
Claude's official document-generation skills were tested but not reliable enough for the strict template requirement.
测试了Claude的官方文档生成技能，但不足以满足严格的模板要求。
The author built an open-source solution after three days of work, accelerated by AI assistance.
作者在AI辅助下，经过三天工作构建了一个开源解决方案。
The solution extracts template characteristics and reuses them faithfully when generating new documents with variable content.
该解决方案提取模板特征，并在生成可变内容的新文档时忠实地复用它们。