SmellCC Wraps SonarQube in an LLM Pipeline for One-Click Python Smell Remediation, Reporting a 96.8% Cleaning Rate at 91.3% Accuracy

arXiv 2608.09477·low signalXBlueskyLinkedInCopy link

Posted 2026-08-10, SmellCC is a VS Code extension that turns SonarQube from a passive alert system into an active remediation tool, using chain-of-thought prompting and few-shot learning to detect and refactor the top-10 most frequent Python code smells in place. The reported quantitative evaluation claims a 96.8% cleaning rate at 91.3% accuracy, with refactored code remaining syntactically correct and behavior-preserving. Worth pairing with the same day's Rope refactoring-oracle result, which found 13 real behavioral bugs in a mature refactoring tool — self-reported behavior preservation is exactly the claim that study shows is hard to keep.