Deploy to agents

A skill or command only helps if your coding agent can find it. yori sync materializes your skills and commands into the directories agents discover them from — rendering templates (vars, includes, slots) on the way, so you compose once and deploy everywhere.

yori sync                          # render into ./.claude (Claude Code, default)
yori sync -a codex -a cursor       # target specific agents (repeatable)
yori sync -a '*'                   # every supported agent
yori sync --global                 # into the agent's global dir (personal)
yori sync --set tone=blunt         # override template variables at deploy time
yori sync --link                   # symlink static artifacts (live editing)
yori unsync -a '*'                 # remove everything sync placed

Where each type lands

Project scope shown; --global uses the personal dir.

Skills carry their bundle support files; a command's {{ input }} becomes the agent's argument token ($ARGUMENTS); a yori agent becomes a Claude subagent with name/description/model frontmatter; a rule keeps its paths: frontmatter (so Claude path-scopes it). Combinations an agent doesn't support are skipped with a note, never an error.

Frontmatter passes through

Any agent-specific frontmatter you author — allowed-tools, argument-hint, agent, context, tools — is preserved on deploy, and runtime syntax (!`cmd`, $ARGUMENTS, @file) is never touched. So you can compose a dynamic command from shared partials and let the agent execute it:

---
name: pr-summary
description: Summarize a pull request
allowed-tools: Bash(gh *)
---
{% include 'house-style' %}        {# yori composes the static scaffolding #}
PR diff: !`gh pr diff`             {# the agent runs this at invocation #}
Summarize for {{ input }}.         {# → $ARGUMENTS #}

yori records what it wrote, so a re-sync prunes artifacts you've removed and refuses to clobber files it didn't create (use --force to override). The deployed skill is a rendered, parameterized copy of your source, not a raw file — the piece a plain installer can't do.

Reproducible setups

yori sync --save records the chosen artifacts to a committed .yori/sync.yaml. A teammate then clones the repo and runs a bare yori sync to hydrate the project's whole agent setup in one command:

# .yori/sync.yaml
agents: [claude-code]
artifacts: [researcher, triage]