Show HN: Claude-计划-审阅者 – 对手 AI 审阅 Claude Code 的计划
2 分•作者: yuu1ch13•大约 2 小时前
Claude Code 有一个计划模式，它会在编写代码之前先写出实现计划。我构建了一个工具，可以拦截这些计划，并将它们发送给竞争对手的 AI（Codex CLI 或 Gemini CLI）进行对抗性审查，然后 Claude 才能继续执行。<p>当 Claude 调用 `ExitPlanMode` 时，会触发一个 `PreToolUse` 钩子，将计划发送给审查者，并返回 `{ permissionDecision: "deny", permissionDecisionReason: "<反馈>" }`。Claude 收到拒绝和反馈后，会修改计划，然后再次尝试。经过可配置的轮数（默认为 2 轮）后，它会通过。最后，你会得到原始计划和最终计划的差异。<p>不同的模型有不同的盲点。其价值在于第二种视角，而不是某个模型“更好”。<p>```
npm install -g claude-plan-reviewer
claude-plan-reviewer setup
```<p>需要 Node.js 18+ 和 Codex CLI 或 Gemini CLI。大约 400 行 JS 代码，零依赖。MIT 许可。<p><a href="https://github.com/yuuichieguchi/claude-plan-reviewer" rel="nofollow">https://github.com/yuuichieguchi/claude-plan-reviewer</a>
查看原文
Claude Code has a plan mode where it writes implementation plans before coding. I built a tool that intercepts those plans and sends them to a competing AI (Codex CLI or Gemini CLI) for adversarial review before Claude is allowed to proceed.<p>When Claude calls `ExitPlanMode`, a `PreToolUse` hook fires, sends the plan to the reviewer, and returns `{ permissionDecision: "deny", permissionDecisionReason: "<feedback>" }`. Claude receives the denial with the feedback, revises the plan, and tries again. After a configurable number of rounds (default: 2), it passes through. At the end you get a diff of the original vs. final plan.<p>Different models have different blind spots. The value is in the second perspective, not in one model being "better."<p>```
npm install -g claude-plan-reviewer
claude-plan-reviewer setup
```<p>Requires Node.js 18+ and Codex CLI or Gemini CLI. ~400 lines of JS, zero dependencies. MIT licensed.<p><a href="https://github.com/yuuichieguchi/claude-plan-reviewer" rel="nofollow">https://github.com/yuuichieguchi/claude-plan-reviewer</a>