review loop skill A reusable AI agent skill that implements an iterative worker reviewer cycle to produce higher quality output through structured critique and revision. Works with any AI agent that supports subagent spawning (Cursor, Claude, Windsurf, etc.). What it does Instead of trusting a first draft, the agent: 1. Does the work (writes code, spec, doc, etc.) 2. Spawns a separate critic subagent with a fresh context 3. The critic scores the work 1–10 with specific, actionable feedback 4. The agent revises based on the feedback 5. Repeats until the quality gate is met (default: 8/10) Result: Fewer bugs, better specs, more robust implementations — without needing a human reviewer in the loop. Quick install [安装/使用命令见下方] Or clone and copy manually: [安装/使用命令见下方] Usage Just mention it in your prompt: [安装/使用命令见下方] [安装/使用命令见下方] [安装/使用命令见下方] Defaults Setting Default Min loops 2 Max loops 4 Quality gate 8/10 Override inline: "use review loop, gate 9, max 5 loops" Folder structure [安装/使用命令见下方] License MIT — free to use, adapt, and redistribute with any AI tool or platform.