What problem does it solve? New owners of a Plow life-assistant agent need a structured first conversation that collects their name, city, sports teams, and calendar selections without forms, repeated questions, or leaking personal setup details into group chats. ## Core Features & Use Cases - Resumable onboarding interview: Reads /var/lib/hermes/ld/config.json each turn and continues from the first missing key, so a paused conversation never re-asks answered questions. - Background calendar discovery: A service snapshots available Google calendars via the Latch relay and renders a numbered pick list, so the owner never types a calendar ID. - Validated config writes: write_config.py deep-merges staged JSON drafts, geocodes the city, enforces a shared config gate, and keeps owner text out of shell commands. - Use Case: An owner texts their new assistant for the first time; the skill greets them, sends the intro sequence, collects city and teams across turns, and offers calendar choices once the Latch snapshot is ready. ## Quick Start Start the onboarding conversation in the owner's one-to-one DM and let the skill collect their name, city, teams, and calendar picks into the household config.
