reflexion:critique

A multi-agent code review eliminates blind spots by orchestrating parallel expert judges to evaluate requirements, architecture, and code quality. You can initiate this by asking the AI to run a critique command on your recent file paths or commits.

Chain-of-verification in a consensus review ensures each specialized judge agent self-validates its critiques before sharing results. This enforces rigorous internal checks prior to the debate and consensus building phase, generating a highly reliable actionable report.

Yes, you can use multi-agent critique for design documents and any deliverable requiring thorough validation. The process accepts file paths, commits, or conversation context, allowing specialized judges to rigorously evaluate architectural artifacts beyond standard code changes.

Specialized judge agents synthesize viewpoints through a structured debate and consensus building process. After running parallel evaluations and enforcing self-validation, they combine differing perspectives into a unified, actionable consensus report for the reviewed work.

The best way to validate requirements and architecture in completed changes is using a multi-agent consensus approach. A Requirements Validator and Solution Architect evaluate the deliverable in parallel, ensuring both design and compliance are rigorously assessed.

You need a parallel multi-agent review when single-perspective blind spots pose significant risks to deliverable quality. This approach is ideal for complex code changes or design artifacts requiring simultaneous validation of requirements, architecture, and code quality.