Agent Grader

For people who built their own

Agent Grader

You put Claude behind Slack, skills on cron, and memory in a database you designed. Ten honest questions tell you what you actually built.

Ten questions · four minutes · no signup

You were right to build

If you built an internal agent system that your team actually uses, you're in the top percentile of the market, and the build itself is proof you read where this is going before most people did. Nobody serious tells a team like yours that you couldn't build it. You could, you did, and you were right to. This grader starts from that — it just doesn't stop there.

Ten questions you should want answered anyway

Almost every DIY build converges on the same shape: a fleet of shared skills, one version for everyone, prompts maintained by the one person who built it, replies that go nowhere. Solo, that works. The walls appear the day a colleague wants their copy to behave differently — every one of these ten questions is a second-user question. Your honest answers tell you whether you built an agentic control plane, or a scheduler with good prompts.

How the grading works

Ten multiple-choice questions. Each has four self-descriptions — pick the one that is uncomfortably, recognizably you. You get a score out of 100, a tier, and a per-question breakdown with what good looks like on every dimension. About four minutes, no signup.