How I Built an AI-Assisted Dependency Vulnerability Fixer

This article details the development of an internal remediation system for dependency vulnerabilities. The author faced 50 open findings across 20 repositories and created a workflow that centralizes findings, classifies fixable issues, and uses GitHub Copilot SDK for AI-assisted code changes. The system validates changes and opens pull requests, balancing automation with human review. It addresses the challenge of reducing engineering time while ensuring safety, covering both mechanical fixes and judgment-required cases. The article highlights the design of a scalable, cross-repository solution that respects tests and ownership.