Linting Style and Substance in READMEs

A team of researchers from the University of Chicago and University of Arizona has introduced LintMe, a novel tool designed to automatically improve the quality of software README files. Presented in a paper conditionally accepted to CHI 2026, the work addresses the critical role READMEs play in shaping first impressions of software projects, noting that requirements vary dramatically—research software needs reproducibility details, while open-source libraries prioritize quick-start guides. The core innovation is a design that moves beyond simple style checkers to handle substantive content issues.

The tool employs a lightweight Domain-Specific Language (DSL) that allows users to create custom, context-specific linting rules. These rules uniquely combine traditional programmatic operations (like checking for broken links) with LLM-based semantic evaluation (such as detecting excessive jargon or missing key sections). This hybrid approach tackles problems that were previously challenging for automated linters. An 11-participant user study found LintMe to be both approachable and well-matched to documentation needs, opening the door for applying similar techniques to more complex documentation and other culturally mediated texts.