AI Skill Grader

Make the skill earn access.

Before you install a skill, you should know what it assumes, how it can fail, and what it can affect.

AI Skill Grader reviews public GitHub skill sources before you run them.

Grade a skill or repo View public method

What we look for

The install question is bigger than “does it sound useful?”

What it claims to do What it can touch Whether it can change things Whether it needs credentials or external accounts Whether instructions conflict with code What to test before relying on it

Question 1

Can I try it safely?

Blast radius: files, credentials, writes, external accounts, sensitive data, paid systems, and safe test paths.

Question 2

Can I trust the output?

Correctness: advice, numbers, code, recommendations, contradictions, and anything you should double-check.

Boundary

A skill can be safe to open and still wrong to trust.

We do not connect to your accounts, use production credentials, or certify that a skill is safe. If something was not tested, we say so.

Testing has to happen on your side. Your data, tools, permissions, model, and workflow can change how a skill behaves.

We show what the public source reveals before install. You still need to test the skill before using it for real work.

Grade a skill or repo