Code review is most useful when it catches defects that automated tools and tests cannot easily identify. A checklist provides consistency, but it should guide judgment rather than force reviewers to comment on every stylistic preference.
Review behavior first
Understand what the change is intended to do and which user workflows it affects. Check edge cases, failure paths, authorization decisions and how invalid input is handled. Confirm that the implementation preserves existing behavior where required.
Review data and security boundaries
Inspect SQL construction, output encoding, file handling, session behavior and access checks. Confirm that database constraints and transaction boundaries support the intended invariants. Check that logs and error responses do not expose secrets.
Review operational impact
Consider migrations, cron jobs, background workers, cache invalidation, configuration changes and resource consumption. A code change can be logically correct but unsafe to deploy if it requires an undocumented sequence.
Keep feedback actionable
Explain the risk and suggest a concrete improvement. Distinguish blocking correctness or security issues from optional refactoring. Prefer small reviewable changes and ask for tests that demonstrate important behavior.
Checklist
- Verify intended behavior and edge cases.
- Review authorization, input and output boundaries.
- Check SQL, transactions and constraints.
- Consider deployment and operational consequences.
- Make feedback specific and proportionate.