I Built an AI-Assisted Dependency Vulnerability Fixer - Here's How I did it

This article details the design and implementation of an internal remediation system for dependency vulnerabilities. The author faced a backlog of 50 findings across 20 repositories and created a workflow that automates the mechanical parts of fixing (like version bumps) while keeping human judgment for risky changes. The system classifies findings, uses scripted or AI-assisted remediation, validates changes, and opens pull requests. The key insight is distinguishing 'mechanically fixable' from 'judgment required' vulnerabilities, and the article emphasizes that even known fixes need validation due to potential side effects. It's a practical guide for DevOps and security engineers looking to reduce toil in vulnerability management.