Graphify agent skill · The Cheddar Chase

Graphify agent skill

Been exploring graphify recently and I think the concept is pretty interesting.

It’s an agent skill that reads a folder of source material, builds a knowledge graph from it, and then gives you a structured view of the concepts, relationships, communities, and key nodes across the whole set.

It gives your agent a reference for a big codebase which stops it from grepping all over and getting lost. What's nice too is there are a lot of optional expansions for non-code use cases as well.

What I like about it:

good for large, messy collections of material
good for mixed inputs like code, markdown, PDFs, screenshots, and diagrams
good for finding structure across a repo or research dump
good for persistent memory, since it saves the graph and lets you query it later
good for agent workflows, especially when you want something more navigable

What I’d avoid using it for:

small codebases where you can already understand everything quickly
situations where you need perfect factual reliability on every relationship
simple “read one file and answer one question” jobs
cases where raw source reading is already cheap enough and the graph layer adds overhead

That’s the part I find most interesting: it feels strongest when the problem is too much material + weak structure.

If you want to check it out head over to: https://github.com/safishamsi/graphify

Curious what other tools anyone else here has tried in this category yet?

3 comments