Loopwise: An automated plan/code review tool. An automated review loop between Claude Code and the Codex CLI. Utilizing two top-tier AI models (Claude Opus 4.6 + GPT-5.4) for cross-review, the output quality is significantly higher than single-model output. V2: Added all the functionality of the official Codex plugin while retaining its original very powerful plan review capabilities. 1. Structured JSON Output — Codex returns structured feedback with severity, confidence, and recommendation. 2. Adversarial Mode (--adversarial) — A skeptical perspective review, checking 7 attack surfaces: authentication, data, resilience, concurrency, boundaries, evolution, and observability. 3. Review Gate (/loopwise-gate) — Quick git diff review before committing. 4. Background Execution (--background) — Does not block current work; use /loopwise-status to view progress. 5. Verify Before Modifying — Claude Code independently verifies each Codex discovery before making changes. 6. Disposition Tracking — Each discovery is marked as verified, dismissed, or unverified_fix. Chinese Documentation: github.com/haohappy/loopwise/b...… github.com/haohappy/loopwise…