se-code-review

Multi-agent code review automates PR analysis by spawning reviewer subagents to evaluate correctness, testing, maintainability, and security. It uses git, gh, and jq to compute diffs and assemble a structured, deduplicated report with suggested fixes.

Yes, standalone current-branch reviews are supported without a pull request. The workflow computes branch diffs using fork-safe base resolution, enabling direct code analysis and structured findings generation.

AI-assisted PR workflow analysis requires bash, git, gh, jq, awk, sed, and grep installed. These command-line tools fetch PR metadata, compute diffs, parse JSON, and assemble the final review report artifacts.

Multi-agent diff analysis covers correctness, testing, maintainability, standards, and security. Multiple reviewer personas evaluate the patch concurrently, consolidating findings into a single deduplicated report with evidence and suggested fixes.

The best way to consolidate AI reviewer findings is to orchestrate subagents that return structured, deduplicated reports. This workflow assembles evidence and suggested fixes into a single artifact ready for PR decisions.

Fork-safe base resolution for branch-based code reviews automatically identifies the correct base commit in forked repositories. It relies on git and gh to fetch metadata and compute accurate diffs across fork boundaries.