Abstract
In large-scale software development, root-cause analysis can be a manual process that may not effectively connect production crashes to nuanced discussions from code reviews. A system for automated software defect analysis can use a closed-loop pipeline to extract engineering principles from historical development artifacts, such as code review comments, to create a structured, machine-readable knowledge base. An autonomous agent may use this knowledge base to analyze new code contributions, generating quantitative risk scores based on principle violations, disparities between developer intent and the code changes, and similarity to historical crash-inducing changes. This process may assist in accelerating root-cause analysis of production failures, and a feedback loop that learns from these failures can improve its ability to identify and help prevent certain high-risk code patterns.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Saha, Snigdha; Li, Zening; and Cheng, Shu-Wei, "Automated Software Defect Analysis and Prevention via a Closed-Loop Knowledge Extraction Pipeline", Technical Disclosure Commons, (September 02, 2026)
https://www.tdcommons.org/dpubs_series/11582