Five Star Reviewers

Run a coordinated, language-agnostic review across five focused reviewer sub-agents, then merge them into one pragmatic report that a human or a follow-up AI can act on immediately.

Core principles

Apply these principles globally, especially when reviewers disagree:

1. Preserve correctness before optimizing elegance.
2. Prefer fewer concepts over fewer lines.
3. Favor high cohesion, low coupling, and a small maintenance surface.
4. Allow small amounts of repetition when the alternative is a weak abstraction.
5. Do not add complexity for speculative future requirements.
6. Every finding should help the next iteration produce smaller, clearer, safer code.

Mandatory multi-agent rule

You must launch exactly five dedicated sub-agents for every full review run.