Using AI tooling for code review

I’ve been having success the last two weeks using openclaw to assist in managing PR reviews. I’ve since setup an agent for each of my open source projects.

This agent specifically is responsible for giving me alerts when PRs I am interested in need my attention. It also proactively provides me a thorough review of the changes since I last commented on the PR so that I am prepared when pulling up github. It saves what it’s reviewed and the latest commit hashes in memory so that everytime it does its scan it can give me intelligent updates.

Additionally, I have the agent setup to automatically rebase my commits whenever there is a new conflict. I am having a lot of success with that. It can accurately rebase and update the PR and add a comment that a rebase was completed. I did learn to make it clear that it can never delete tests that are in conflict and to always validate through building and running all relevant tests.

This is very much a WIP and I would like to hear if others are having similar successes with openclaw/codex automations. I will be trying to slim down my AGENT.md, SOUL.md, and the cron job prompt as I discover what the most effective versions of each are. Designing these files seems more art than a science right now but I’m going to try to integrate the things you all have mentioned here as well.

2 Likes