This plugin enhances the Claude Code CLI by introducing an automated, two-phase code review loop. It targets developers using Claude Code who seek an independent, AI-driven second opinion on their code changes before integration. The primary benefit is ensuring code quality, consistency, and adherence to best practices through an automated review process integrated directly into the development workflow.

How It Works

The plugin operates via a two-phase lifecycle triggered by the /review-loop command. Initially, in the "Task phase," the user describes a coding task, and Claude implements it. Upon completion, a "Stop hook" automatically initiates the "Review phase." This hook spawns up to four parallel Codex sub-agents—Diff Review (code quality, tests, security), Holistic Review (structure, docs, architecture), Next.js Review (framework specifics), and UX Review (UI, accessibility)—tailored to the project type. These agents analyze the changes, and their findings are deduplicated and consolidated into a Markdown file (reviews/review-<id>.md). Claude is then prompted to address the feedback, ensuring issues are resolved or acknowledged before the loop concludes.

Quick Start & Requirements

• Installation:
  • From CLI: claude plugin marketplace add hamelsmu/claude-review-loop followed by claude plugin install review-loop@hamel-review.
  • From Claude Code session: /plugin marketplace add hamelsmu/claude-review-loop then /plugin install review-loop@hamel-review.
• Prerequisites:
  • Claude Code (CLI)
  • jq (macOS: brew install jq, Linux: apt install jq)
  • Codex CLI (npm install -g @openai/codex)
• Configuration: The Codex multi-agent feature is automatically enabled in ~/.codex/config.toml on first use.
• Timeout: The stop hook has a default timeout of 900 seconds (15 minutes), configurable in hooks/hooks.json.

Highlighted Details

• Parallel Review Agents: Leverages up to four specialized Codex agents running concurrently for comprehensive analysis.
• Multi-faceted Review: Covers code quality, test coverage, security (OWASP Top 10), project structure, documentation, Next.js specifics, and UX/accessibility.
• Automated Feedback Loop: Claude automatically addresses consolidated review findings, blocking exit until feedback is handled.
• Consolidated Output: All findings are deduplicated and presented in a single reviews/review-<id>.md file.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or roadmap were found in the provided README.

Licensing & Compatibility

The README does not specify a license. Compatibility for commercial use or linking with closed-source projects cannot be determined without a license.

Limitations & Caveats

The default Codex flags (--dangerously-bypass-approvals-and-sandbox) may present security considerations; a safer --sandbox workspace-write option is available. The 900-second stop hook timeout might be insufficient for very large or complex code reviews. The plugin relies heavily on the Claude Code CLI and Codex CLI infrastructure.