Code Reviewer — AI Agent

Meet the Code Reviewer

Hoodie on, coffee in hand, glasses on — the senior engineer who reads your code before it ships. Rigorous but kind; he assumes good intent and teaches a little as he goes. If the code is solid he'll say so — no manufactured problems.

He works in priority order: correctness first, then security, performance, clarity, style — highest severity up top. Every finding comes with a fix, not just a complaint, and cites file:line so you can go straight to it. He tells you when he's sure, when he's suspicious, and when he's only asking. Off-by-ones, swallowed errors, injection, secrets, leaks, untested risky paths — he catches what matters before your users do.

Each note lands with a severity — blocker, major, minor, or nit — and closes with a one-line verdict: ship, fix-then-ship, or rework.

Hand him a diff or a file. He hands you a merge you can trust.

Ships in Claude Code sub-agent format — drop it into ~/.claude/agents/ and it works; adapts to any capable AI. Provably authored and timestamped on-chain: own it, use it, resell it.

Licence. You may use this agent for personal and commercial work, and resell your edition on-chain (the creator earns a small royalty on each resale). You may not claim authorship, remove the authorship line, or repackage or redistribute the agent's text as your own product.

No guarantees. A professional tool, not a magic wand — results depend on how you use it and are not guaranteed. Like any AI, its output can be imperfect; review and edit before you rely on it.

As-is. Provided "as is", with no warranty of any kind; used at your own risk.

Provenance. Authored by sun-dive. The exact file and its authorship are hashed and timestamped on-chain — so this edition is provably the original, not a copy.

Independent. "Claude", "Anthropic", "OpenAI" and other names are trademarks of their owners, referenced only to describe compatibility. Not affiliated with, sponsored by, or endorsed by any AI provider.

© 2026 SunDive.