ASD-STE100 Simplified Technical English for LLMs (github.com)

A new release focusing on Simplified Technical English (STE) for Large Language Models (LLMs) has emerged, featuring the ASD-STE100 agent skill that allows for two operational modes: strict and flavored. This tool promises a self-contained approach, needing only Python 3 to function, and comes equipped with a heuristic linter (ste-lint.py) designed to reduce errors in technical writing. It employs a measurable scoring system to analyze drafts, offering feedback that can significantly enhance clarity and quality by lowering the number of writing violations. The significance of this development lies in its ability to streamline writing for both human users and LLMs, as evidenced by experimental results comparing models like Claude and GPT-5.5 across various writing tasks. Notably, the introduction of STE reduced "slop"—the term for unclear or imprecise writing—by over 50% in many test scenarios, outperforming other methods, such as banned-words lists. While this tool cannot fully certify compliance with STE guidelines (which still require human judgment), it provides a promising mechanical foundation for enhancing the quality of technical communication in AI applications. Overall, this advancement supports the AI/ML community by offering a practical way to improve the effectiveness and readability of machine-generated text.