parallel-code-review

Automate pre-merge code review by snapshotting the origin/main to HEAD diff and dispatching parallel Sonnet reviewers. This aggregates security, dependency, and test evidence into a single BLOCKED, WARN, or OK verdict to halt risky merges.

Parallel code review partitions a narrowed code diff into disjoint file-scope chunks while preserving seam-context nuclei. Multiple reviewers analyze these chunks simultaneously, and a synthesizer emits a structured BLOCKED, WARN, or OK verdict for release gating.

Run a security audit and CVE dependency scan by executing a weekly code-review gate that dispatches dedicated security and dependency workers across the full diff. This validates findings on disk before synthesizing them into a final merge verdict.

No, you cannot trigger a merge gate review outside the weekly cadence. The review process enforces strict merge-cadence rules and explicitly refuses to run outside the designated /workweek-complete cycle to prevent out-of-cycle gate misuse.

Yes, parallel code review works without rewriting source files by using a no-rewrite synthesizer. It aggregates evidence from multiple reviewers and outputs structured BLOCKED, WARN, or OK verdicts alongside evidence outputs, leaving the codebase untouched.

The best way to structure code chunks for parallel review is using seam-first partitioning. This builds disjoint file-scope chunks while preserving seam-context nuclei, which improves convergence and avoids duplicated review coverage across the diff.