Life OS — Data skill for Claude Code

Life OS


A tiny personal-dashboard framework: **one data layer on your own PC, three ways to touch it** — an offline-first Android app, browser dashboards with inline edit controls, and your AI coding assistant. No cloud, no accounts, no subscriptions; Python standard library on the desktop, Kotlin/Compose on the phone, SQLite files you own.

The design premise: you should not build dashboards by hand in a no-code tool. You describe a life domain to a coding agent (Claude, or any capable assistant), point it at `PLAYBOOK.md`, and it writes a purpose-fit dashboard — schema, sync, HTML, and optionally native phone screens — in an afternoon. The framework supplies the proven plumbing so the agent only writes the domain.

The reference dashboard, **casework**, tracks a day in the life of a modern consulting detective — open cases, leads, and contacts for a certain resident of 221B Baker Street. All of its data is fiction; replace it with your own domains.

What's in the box

path	what
server/sync_server.py	the one desktop server: static dashboards, phone sync API, browser-edit endpoints
server/dashboards.config.json	engine registry — add a line per dashboard
server/dashboards.json	the card manifest the app + hub render (no rebuilds to change it)
server/home.html	browser hub: every dashboard in tabs
example/casework/	the reference engine + generated dashboard + seeded fictional data
app/	Android app template: offline-first editing (Room + outbox + WorkManager), manifest-driven cards
PROTOCOL.md	the sync contract (pull cursors, idempotent push, tombstones, conflict policy)
PLAYBOOK.md	the add-a-dashboard checklist your coding agent follows

Quickstart

**Windows note**: clone with `git -c core.longpaths=true clone ` — the app template's screenshot-test goldens have long generated filena