Spins up three parallel reviews when you type /review: one critic checks code quality and patterns, two plan reviewers assess architecture alignment and change impact, then a synthesis agent combines everything into a verdict (approve, request changes, or needs discussion). The parallel execution is smart since reviews don't depend on each other, so you get comprehensive feedback faster than sequential passes. Works for PR reviews or just sanity checking before you merge. The verdict structure with blocking versus non-blocking issues is genuinely useful because it tells you what actually needs fixing versus what's just polish. You can run quick mode for just the critic if you want fast feedback without the full workflow.
npx skills add https://github.com/parcadei/continuous-claude-v3 --skill review