claude-code-review-council

Multi-agent code review by delegating to (1) codex-cli with gpt-5.6-sol high, (2) gemini-cli with Gemini 3.1 Pro, and (3) five Claude specialist subagents (security, performance, logic, regression, robustness — the last one run blind, without the intent briefing) spawned via the Agent tool — then synthesizing all seven streams with your own diligence. Use this whenever the user asks for a deep review, second opinion, council review, multi-tool review, or wants the most thorough analysis of code changes / a PR / a commit / staged changes / a branch — even if they don't explicitly say "review-council". Stronger than any single-model review because different model families catch different classes of issues and the specialist subagents add focused depth on each axis; the synthesis step verifies findings against the actual code before applying fixes.