review-loop

Multi-agent code review automation dispatches a review loop over your session work to surface and verify PR diff changes. It coordinates deterministic lint, test, and build checks alongside a falsifier stage until the diff is clean.

A falsifier stage in automated code review attempts to actively disprove or break the changes in your diff. It works alongside deterministic lint and test checks within a multi-agent loop to verify code changes before recording a commit-pinned verdict on open PRs.

Yes, you can enforce an auditable PR verdict trail automatically by invoking the review loop through a Stop hook. This dispatches the multi-agent review process and records a commit-pinned verdict on open PRs when the session ends.

You can run a re-review loop on code changes by invoking the review automation manually or via a Stop hook. It iterates through lint, test, build, and falsification stages repeatedly until the diff is completely clean or the review budget is exhausted.

No external dependencies are required to run the automated multi-agent review loop. It independently coordinates deterministic lint, test, and build checks alongside a falsifier stage to verify your diff without relying on additional packages.