Okay, I have to share this little weekend experiment that's already saving me from some AI-induced headaches. We all love Windsurf's speed, but sometimes its code suggestions, especially for nuanced style rules or deprecated patterns, need a final sanity check, right?
I built a simple VS Code extension that intercepts Windsurf's accepted output and pipes it through a local linter *before* it hits my codebase. It's basically a middleman. So, if Windsurf suggests a React component using an old lifecycle method, my ESLint config (which is pretty strict) catches it immediately, and I see the warning inline.
The setup is surprisingly straightforward. The extension:
* Listens for document changes triggered by Windsurf's accept action.
* Runs a quick lint on the changed section (or file).
* Surfaces any warnings or errors directly in the editor, without blocking the flow.
It works with any linter you have configured (ESLint, Pylint, Rubocop, etc.). The key is it's passive—it doesn't stop the suggestion from being applied, it just flags the linter output right after. This means I can still accept a suggestion quickly, but with an immediate visual cue if something's off.
It's made me much more confident about using "Accept All" in some scenarios. Has anyone else tried something similar? I'm curious if others have built little bridges between their AI coding tools and their quality-check systems.
✨ laura
null