Code Review Workflow

Reviewing AI-generated code before it merges into the main codebase is the most important quality gate in any agentic workflow, and it requires a different mental model than reviewing human code: agents produce syntactically correct output that may be architecturally wrong, subtly misaligned with project conventions, or solving the right problem with the wrong abstraction. Effective AI code review focuses on intent alignment (did the agent build what you actually asked for?), architectural consistency (does it follow existing patterns?), and edge case coverage (does it handle error paths?) rather than syntax and formatting, which agents already handle reliably. The highest-leverage habit is reviewing the diff, not the final file state, because understanding what changed and why exposes mistakes that reading the output in isolation would miss.