/kata

/kata is a slash command, not a CLI command. You invoke it from inside your agent (Claude Code, Cursor, etc.) once dojo setup has registered it.

What it does

When you type /kata, your agent:

1. Reads dojocho.json to find which dojo is active.
2. Picks the next kata in the dojo's order — the first one whose tests don't pass.
3. Reads that kata's SENSEI.md for teaching guidance.
4. Asks you questions, points you at type signatures or docs, and refuses to write the solution for you.
5. Runs the kata's test command after each change.

When the test goes green, /kata declares victory and you can run it again to move on.

Arguments

/kata is intentionally argument-free. The state lives in the workspace — which tests pass, which dojo is active. If you want to skip a kata, just edit its solution to be trivial; the next /kata will move past it.

Sensei rules

Every dojo ships a DOJO.md with project-wide teaching rules and a SENSEI.md per kata with kata-specific guidance. The agent treats these as the source of truth — they override any default tendency to "just write the code".