review-recent-work-subagents

Automated multi-agent code review orchestrates specialized subagents to inspect contract correctness, regression surfaces, and lifecycle boundaries across touched files, then synthesizes and applies a verified fix plan.

A multi-agent review workflow spawns six wave-1 subagents to examine distinct quality dimensions, synthesizes findings into a provisional fix plan, and validates residual gaps with two wave-2 closure reviewers before implementation.

Subagents can independently check contract correctness, failure modes, state lifecycle, ownership, and validation boundaries while the parent agent remains read-only and collects evidence to synthesize fixes.

Post-implementation validation is coordinated by a parent agent that spawns subagents to inspect touched files, synthesizes a provisional review-and-fix plan, and guides implementation before re-running verification to report results.

Multi-agent code review requires no external dependencies or components, operating entirely by orchestrating specialized subagents within the existing agent framework to validate recently implemented ExecPlans.

A two-wave review process uses wave-1 reviewers to gather material code-grounded findings across distinct dimensions, then deploys wave-2 closure reviewers to validate residual gaps before producing a tight, implementable fix set.