[Feature Request] Implement mandatory human code review process for AI-generated changes

Bug Description
I understand you want to let Claude Code do all of the actual coding of Claude Code, but it is becoming increasingly buggy as a result. The sheer number of bug fixes every day is not a good thing -- to me it points out that AI slop is creeping into the code base and you can no longer review what it is doing. There are a number of fundamental bugs I have reported that have gone unfixed for weeks. Yet you seem to be more interested in adding dozens of new features, presumably all written by Claude with little or no oversight.

My experience in the last few weeks has shown that even running a thorough review of plans and then /simplify on the code does not catch everything, or there are multiple decisions that must be made by me (an actual human), or I can suggest a better solution. Saying "the tests pass so it must be okay" is like saying, "The goal is to get to LA from SF... and I don't care how I get there, even if it's by way of NY."

Environment Info

• Platform: darwin
• Terminal: iTerm.app
• Version: 2.1.70
• Feedback ID: 9f99feb0-e7b1-4925-8f2a-4ab7eff3ed3c

Errors

[]