Useful automated code reviews by AI
modules: general purpose rules to focus onspecifics: things specific to your team or projectrepos: clone all the repos you care about in here
With Claude:
review <branch-name> in <repo-name>
- Diff the branch against
origin/main - Evaluate in parallel against each module (general best practices)
- Evaluate in parallel against specifics (team/project rules)
- Generate a summary of potential issues
- Fact-check each issue against the actual codebase to eliminate false positives
- Output PASS/FAIL per module with locations and suggested fixes for failures