setup-playbook-jupi
setup-playbook-jupi — bootstrap the declared world
One attended run that takes a workspace from zero to ready to run the process: the owner interviewed (always — whatever they had written down), the playbook extracted from their documents and from the one the interview writes (holes included — on day 1 the playbook is mostly declared ignorance, and that is the point, §2), the dossiers created from the declared source, the human-readable projection rendered. The playbook is the prior, decisions are the evidence, rules are the posterior (§1).
Posture (proactive-setup parity): narrate every step — open it with what you're about to do, close it ✅ done / 🔧 fixed / ⚠️ needs you. Front-load everything human-gated into the prelude; after the ✋ boundary nothing may prompt. If nobody answers a prelude question, ask once, save what is settled, and halt — a re-run resumes.
Frame every step as work being lifted off their plate, never as configuration. The user answers questions about their own work, in their own words; you write every file. Never ask anyone to create, open, copy or fill a config file — and never ask for a value you could discover, derive or default (a path you can find, a stage you are about to declare yourself, a threshold with a sane default). A question that only a human can answer is worth asking; every other question is a bug.