dojo setup
One-time wiring of a project so your agent can run katas.
Usage
dojo setupdojo setup prepares the current directory to host one or more dojos. It is
idempotent — running it twice is harmless.
What it does
- Detects which coding agent you're in via env vars (e.g.
CLAUDECODE=1for Claude Code) and registers the/kataslash command there. - Writes the sensei rules — the agent-side instructions that enforce Socratic teaching: ask questions, never spoil solutions, narrow scope to the next failing test.
- Creates the expected workspace layout so installed dojos drop in cleanly.
Specifying the agent explicitly
If dojo setup can't detect your agent, or you want to wire up multiple, pass
one or more flags:
dojo setup --claude # Claude Code
dojo setup --opencode # OpenCode
dojo setup --codex # OpenAI Codex CLI
dojo setup --gemini # Gemini CLI
dojo setup --pi # Pi
dojo setup --claude --codex # multiple at onceRe-running setup
Re-run dojo setup if you've upgraded the CLI, switched agents, or want to
reset the agent rules to defaults. Your code is untouched.