workflows-review

Multi-agent code review is automated by coordinating isolated git worktrees to analyze diffs between the target and default branches. This process surfaces code quality issues, security concerns, and architectural risks while generating actionable TODOs.

AI code review can target specific files or the full diff between the default and target branches. The multi-agent analysis evaluates the selected scope to identify and summarize security concerns and structural risks.

Git worktree isolation improves AI code triage by keeping the review environment clean and reproducible. By evaluating branches in isolated worktrees, parallel agents can safely analyze diffs without interfering with your primary working directory.

The best way to triage architectural risks in a pull request diff is using ultra-thinking synthesis to evaluate the changes. This approach surfaces deep structural concerns across the target branch and outputs structured recommendations.

No special dependencies are required to perform automated security reviews, but the environment must support git worktrees. The skill enforces safety by outlining required tooling and producing reproducible output artifacts like todo files.

A limitation of using parallel agents for code analysis is that the evaluation is strictly scoped to the diff between the default branch and the target branch, or specifically selected files, rather than evaluating the entire repository context.