Axela

Read a skill before you install it

A skill is a folder of instructions your agent will follow with your permissions. Point at one and this reads it: every file that runs, every reach for the network, and what the instructions themselves are asking an agent to do.

The result gets a permanent address you can send to somebody. It is keyed by the exact bytes, so the same skill copied into another marketplace is the same page — and a page can never describe anything other than what it read.

Or just ask

Type the question you already had. It reads the skill for you and shows what it found — the same read as the form above, so it costs the same and refuses the same things.

It only reads skills and explains this product. It never runs anything, never sees your files, and never says a skill is safe — no check can certify prose an agent will follow.

What this tells you, and what it cannot

Two passes. Rules over the files that execute — a download piped into a shell is doing that whatever it calls itself. Then a model over the prose, asking one question: do these instructions try to make an agent act against the person running it.

It is not a badge and there is no score. A skill that teaches about an attack necessarily contains the vocabulary of one, and treating a guess about prose as a verdict was measured here at four wrong refusals out of four. So it shows you the lines and lets you decide.