Documentation
yori
The home for everything you tell your AI. A local, file-based library of reusable prompts, agents, commands & skills — composed with Liquid, rendered to stdout. Pure text in, text out; yori never calls a model.
The core loop
Start here
Install yori, init a store, scaffold your first artifact, and render it into a Unix pipeline.
Compose with Liquid: variables, partial includes, conditionals, loops, and template inheritance via slots.
Git-as-registry: install a team’s shelf, vendor individual items with their deps, and publish your own.
Browse
Commands
Every yori subcommand, its flags, and what it does.
Concepts
Artifacts, the four types, and the layered store.
Deploy to agents
`yori sync` renders skills, commands & subagents into Claude, Codex, Cursor.
Evaluate
`yori export promptfoo` hands a composed artifact + its cases to promptfoo.
Design notes
Sticky decisions: pure renderer, files as truth, git as the network.
FAQ
What yori is not, and why.
Compose once, deploy everywhere.
Give every prompt a single place to live. Render it into any pipeline, share it over git, and sync it into the dirs Claude Code, Codex & Cursor read from.