Quality Audit

After implementation, the Reviewer agent performs a read-only audit against the approved design. It checks architecture compliance, verification coverage, security-sensitive areas, and code quality. Blocking findings pause the workflow until they are resolved or explicitly accepted.

How it works

The Reviewer agent compares the implemented code against the specification, checking every acceptance criterion. It evaluates architecture compliance (correct patterns, proper abstractions), reliability (error handling, test coverage, edge cases), security-sensitive surfaces, and code quality (naming conventions, complexity, duplication). The result is a concrete finding list with blocking versus advisory issues. Critical findings block progression -- the Builder must fix them before the workflow can continue.

Quality audit report showing blocking and advisory findings

Screenshot coming soon

Why it matters

Human code review is expensive and inconsistent. Different reviewers catch different things, and time pressure often leads to rubber-stamped approvals. A structured audit provides a repeatable quality baseline for every change. It does not replace human review for architectural decisions, but it eliminates the class of issues that come from inattention: missing error handling, incomplete tests, naming inconsistencies, and spec drift.

Before and after comparison of workflow audit quality

Screenshot coming soon