Code review
Review bots apply the vendor's idea of a good change. Your real rules (fail closed on auth, two-person review on payments, skip generated code) live in people's heads.
How it runs
Your review standards, run by your agents on every pull request.
The diff is read with your coding standards, Skills & Rules, and the review decisions your team has already made.
Where it stops
A person accepts the finding, or records why it does not apply. Sensitive paths always wait for one.
What the next run learns
A waiver becomes context for the next review, not a lost Slack thread.
Also in Engineering
Run it on your own agents.
Start from a starter or build it your way, with the gates and limits you choose.