Audit provides a multi-level code review for recent changes, evaluating quality, specifications, security, and performance across levels L1 through L5. It gathers context by mapping surfaces, indexing semantics, and scanning conventions before dispatching domain specialists for targeted reviews. Findings are synthesized and reconciled, culminating in a fix gate that can automatically apply recommendations. Reach for it when you need a rigorous, multi-agent code evaluation on a pull request or commit range.