一款开源Agent Skill工具被开发用于约束大语言模型按照ASD-STE100简化技术英文标准编写文档[1]。该标准是航空航天业自1983年以来采用的受控语言规范，包含53条编号规则和9个章节，目的是确保文档的清晰度和准确性[1]。
该工具经过测试验证了显著效果[1]。在对6个Claude模型进行的96次运行测试中（涵盖8项任务和2种条件对比），STE违规率平均下降72.9%[1]。同时，生成的文档在所有模型上的输出token数量均有减少，平均句子长度从11.2词缩短至9.7词[1]。该工具采用MIT开源许可证发布，兼容Claude Code、Cursor、VS Code Copilot、OpenAI Codex、Gemini CLI等约25种AI工具[1]。
An open source Agent Skill tool has been developed to enforce compliance with ASD-STE100 simplified technical English standards when large language models generate documentation [1]. ASD-STE100 is a controlled language standard that has been used across the aerospace and defense industry since 1983 to ensure documentation clarity and accuracy [1].
Testing of the tool demonstrated significant improvements in documentation quality across multiple AI platforms [1]. The skill reduced STE violations by 72.9% across six Claude models while simultaneously decreasing output token consumption [1]. Average sentence length was reduced from 11.2 words to 9.7 words [1]. The testing evaluated performance through 96 runs spanning six Claude models, eight different tasks, and two condition sets [1].
The tool is compatible with approximately 25 different AI platforms and development environments, including Claude Code, Cursor, VS Code Copilot, OpenAI Codex, and Gemini CLI [1]. The Agent Skill incorporates the full ASD-STE100 standard, which comprises 53 numbered rules organized across 9 chapters [1]. The project is released under an MIT open source license [1].