First deploy

Build this on cubek. Install the CLI (`curl -fsSL https://cubek.dev/install | sh`), then run `cubek skill` and read all of it before writing code; follow it exactly (if this project already exists, port it the way the Skill says).
No CLI? Download the Skill verbatim (e.g. `curl -fsSL https://cubek.dev/SKILL.md`); don't summarize it.
Sign me in with `cubek login`.
Run `cubek check` until it is clean, then `cubek deploy --prod`.
Reply with the live URL.

The prompt above is all your agent needs: it reads the Skill, builds, checks and deploys, and replies with the URL. The same steps by hand:

cubek init my-app
cd my-app
cubek dev

cubek dev serves the app at http://127.0.0.1:8787 and recompiles on save. After every edit, cubek check compiles the app and prints each error with its canonical fix; cubek test runs the *.test.ts files.

cubek login

The browser opens the sign-in page; sign in with GitHub or Google. Projects go to your personal organization; cubek org switch <slug> picks another. In CI, use cubek login --token T with a personal token or a tenant API key, or set CUBEK_URL and CUBEK_KEY.

cubek deploy          # a preview: https://my-app--r1.cubek.app
cubek deploy --prod   # production:  https://my-app.cubek.app

Every deploy is an immutable revision with its own preview URL. --prod also snapshots the database, applies additive migrations and switches traffic atomically. Destructive schema changes need --allow-destructive.

cubek revisions            # * marks production
cubek rollback 3           # back to r3 (add --with-db to restore its data)
cubek logs --follow
cubek sql "SELECT COUNT(*) FROM notes"
cubek secrets set STRIPE_KEY sk_live_...
cubek db restore --to 2026-09-26T10:00:00Z