Reviewer
The Reviewer agent performs the post-implementation audit. It checks architecture compliance against the specification, evaluates reliability through test coverage and error handling, and assesses code quality including naming, complexity, and pattern adherence. Critical findings block the workflow until resolved.
How it works
The Reviewer reads the specification and the implemented code side by side. It checks each acceptance criterion against the test results. Architecture is evaluated for pattern compliance and proper abstractions. Reliability covers error handling, logging, timeout management, and test comprehensiveness. Code quality examines naming conventions, cyclomatic complexity, duplication, and adherence to project patterns. The output is a concrete finding list with severity and suggested follow-up.
Quality audit with specific findings and severities
Screenshot coming soon
Under the hood
The Reviewer maintains a checklist of project-specific quality criteria derived from CLAUDE.md and vault knowledge. It does not just check for generic code quality -- it understands your project's specific patterns and conventions. Critical findings (security vulnerabilities, missing error handling on external calls, untested acceptance criteria) block the workflow. Advisory findings are noted but do not block.
Quality gate decision matrix showing blocking vs advisory findings
Screenshot coming soon
Related Features
Quality Audit
A reviewer agent audits implementation against the original design for Architecture, Reliability, and Code quality.
Builder
Test-first implementation: writes tests before code, verifying every requirement.
Security Officer
Deep security scanning, threat modeling, vulnerability detection, and gate authority.