ce:review

A multi-agent code review works by spawning parallel reviewer personas to inspect PR diffs and return structured JSON findings. These findings are merged, deduplicated, and staged with confidence gating to deliver a comprehensive report including risk notes and testing gaps.

Automated code review supports branch diffs and standalone checks alongside standard PR reviews. The workflow applies to any code changes, generating structured JSON findings with confidence metrics, risk notes, and testing gaps for comprehensive intent discovery.

Multi-agent PR review supports autofix-safe automation and report-only reviews to enable automated quality gates. It produces structured JSON findings with confidence metrics, allowing you to gate code merges based on risk notes and testing gaps identified by the reviewer personas.

The output format for structured code review findings is JSON, containing confidence metrics, risk notes, and testing gaps. The system optionally surfaces project-standards paths for governance, merging and deduplicating results from parallel sub-agents into a single report.

Diff analysis identifies testing gaps and risk notes by spawning dynamically selected reviewer personas to inspect code changes. These parallel sub-agents return structured JSON findings which are merged and deduplicated, applying confidence gating to highlight comprehensive intent discovery and project-standards governance.