The hardest part of review tools like this is signal-to-noise — one round of false positives and people start reflex-accepting or muting it. How does Ox decide what's worth flagging on a given PR, and does it dedupe against pre-existing debt so it only surfaces what the change actually introduced? In my company, we use CodeRabbit with Azure DevOps and it flags a lot of things and mostly they were false positives so eventually we stopped relying on it.

Existing debt is currently handled in a debt queue structure. If you ignore a suggestion once it will remember that and store it in that queue so that it doesn't bring it up again. There is a mechanism to review that queue but I haven't totally thought out how to resurface those in a convenient manner.

There are obvious solutions like dumping those ignored changes to Jira issues so you have a pipeline of debt to address though I feel like this just moves the issue so I'm trying to think of something more proactive.

I had same issues as such tools provide noise more than something concrete.