I was pairing with an AI assistant on my own codebase and realized I was re-explaining the same things every single session. The same conventions, the same gotchas, the same reason we do that one thing the weird way. Brilliant at generic code, useless at mine. The problem: everyone treats this as a prompting problem. It isn't. When a new engineer joins you don't write them a better prompt — you point them at the docs, the past decisions, and the code that shows how things are done here. Agents need the same thing, in a form they can load. How it evolved: I started with "turn a doc into a skill". That worked, but it still needed someone to write the doc. Then it clicked that the knowledge was already sitting in the repository. Git Timeline reads your project's history, renders the year as a short film, and writes a skill from the patterns it finds — the practices your team already follows but never wrote down. That's the part I didn't plan and now like most. One honest note: the evidence on AI tooling cuts both ways. METR measured experienced developers 19% slower with AI while they felt 20% faster. I think that gap is missing context rather than missing capability, but I'd rather put it on the table than pretend. Free to try at loreto.io — point it at a repo you know well and tell me whether the skill it writes is actually right. That's the feedback I want.

Loreto

Turn any repo, doc, or video into an AI agent skill

Turn any repo, doc, or video into an AI agent skill