Multi-lens code review with planner-driven depth selection, 6 review lenses, AI failure-mode detection, finding verification, and evidence-quality discipline. Produces severity-ranked findings with confidence scores and gates PR creation on zero blocking issues.
