Audit web↔iOS consistency for hand-copied surfaces (preset tables, metrics-catalog, debrief taxonomy, API DTOs) and surface divergences as an actionable task list. Run after a feature that touched one platform, before merge, or any time the two clients may have drifted.
Rsync the latest complete ECMWF GRIB run from the production server into the local ECMWF_GRIB_DIR so refresh can run with ECMWF enrichment locally
Deploy the weatherbrief app to production on weather.flyfun.aero
Start or restart the local dev server (backend + frontend) in a per-worktree tmux session. Use --https (or --simulator) to run the singleton TLS instance for iOS simulator testing.
Run LLM digest eval — replay saved weather contexts through the LLM and compare assessments
Build & maintain the LLM-digest eval SET (corpus) — pull prod briefings, attach pilot debriefs, re-run advisories against saved baselines, promote staging→corpus. Distinct from the `eval-digest` skill (which replays contexts through the LLM).
End-to-end production health check for weatherbrief on the flyfun.aero droplet — droplet metrics, container state, log signals, refresh queue, standalone cycle, RSS growth
Create a new git worktree with its own venv, dependencies, .env, and shared data dir. Invoke with the branch name as the only argument.