Any data in. A schema your agent can actually use, out.

How it works

Most ingestion tools summarize: read everything, compress, done. skemify works the way a curious expert works instead — asking, reading, and writing up in a very particular order.

AIM

Before it reads a word, it writes down what it’s actually trying to find out about your data — what would matter most if it turned out to be true, what would change everything if it turned out to be false.

RANK

It works those questions in order of importance, not in whatever order your files happen to be in.

SHOOT

It reads in scoped bursts, predicting what a section will say before it opens it, and keeps only what surprised it or proved a prediction wrong. A confirmed guess doesn’t need retelling.

SWEEP

Every so often it stops and asks what none of its questions caught — the stuff nobody thought to look for is often the stuff that matters most.

SKIN

It writes the result back out backbone-first: the handful of things that matter most, with everything else hung underneath them — never a flat transcript of everything it touched.

Every piece of your source material is accounted for on the way through — placed somewhere in the structure, deliberately dropped, or logged as an open gap. Nothing just vanishes.

What you get

A graph: nodes with a why-this-matters line, and edges weighted by how load-bearing each connection is, from 5 (miss this and everything downstream breaks) to 1 (nice to know). Alongside it, an AGENTS.md file that teaches your own agent the same reading order: start at the backbone, follow the heaviest connections first, and treat a logged gap as a gap rather than silence.