Copy-paste prompts for reviewing AI-generated code in 4 rounds. Free and open source. Star it, fork it, add your own.
AI-generated code looks correct. But it fails in ways human code doesn't.
This repo contains the copy-paste prompts and an optional automated review skill from The AI Code Review Protocol. Use them as-is, or fork and adapt for your team.
| Round | Prompt files | Use for |
|---|---|---|
| 1 — Surface Scan | 1.1, 1.2, 1.3 | Logic errors, wrong assumptions, edge cases |
| 2 — Security Deep-Dive | 2.1, 2.2 | Injection, authorization, CSRF, attack paths |
| 3 — Architecture Smell Check | 3.1, 3.2, 3.3 | API correctness, codebase fit, over-engineering |
| 4 — Comparison Pass | 4.1, 4.1b, 4.2 | Requirements trace, scope drift, user journey gaps |
A copy-paste agent skill is included at:
Use it when you want an agent to run Rounds 1 and 2, apply minimal fixes, re-scan, and then stop for human judgment on architecture and requirements.
This repo is a free companion to The AI Code Review Protocol ($8.99/ZAR149 on Selar).
The full guide explains why each round exists, traps to watch for, when to reorder rounds, how to run the review loop, and how to adapt the prompts into your own automated workflow.
By Stephen Metcalfe (@raithlin)