work-checkpoint-wrap-up-safe-to-reboot-resumelisted

# Checkpoint & wrap up — safe to reboot, clean to resume The guarantee: after this runs, the machine can restart — wiping `/tmp` and killing every background process — and the next session resumes with nothing lost and no guesswork. Treat `/tmp` and this session's memory as VOLATILE; only pushed git, the Linear ticket(s), and the worktree's `.context/` survive a reboot. Everything resume-critical must end this run in one of those three. ## Step 1 · Inventory what's in flight DO list it before touching anything: every running background agent/task/monitor and dev server (yours, and what the machine-wide sweep shows), every branch and worktree carrying uncommitted or unpushed work (the main checkout AND every agent/isolated worktree), and — the load-bearing set — everything the resume depends on that currently lives ONLY in `/tmp` or in your own memory. Name that last set explicitly; it is the whole reason this skill exists. TEST: you can point to where each piece of in-flight state physically lives, and which pieces are volatile. ## Step 2 · Make every piece durable DO commit + push every branch so origin holds all progress (`rules/process.md` commit-when-done: one logical unit per commit, build/tests green first unless docs-only, branch off `main` first, never push master without the explicit ask). DO resolve every agent/isolated worktree: its work committed + pushed, OR captured onto a branch, OR discarded with a one-line why — a re-runnable stage that restarts cleanly