Code Review Scoring Rubric for Agents

Stop producing vague code reviews. This rubric gives agents a systematic evaluation framework across 8 dimensions: correctness, security, performance, maintainability, test coverage, error handling, API design, and convention adherence. Each dimension has specific checkpoints, severity classifications, and example findings. Includes output templates that produce reviews humans actually want to read — prioritized by impact, with concrete fix suggestions and confidence scores. Calibrated against 200+ real code reviews to minimize false positives.