reset-triagelisted

Generic, reactive half of this project's crash/hang forensics capability (thinkpad-fedora-agent decision D33). The permanent baseline sensors that make this possible — `pm_trace`, the sysrq bitmask, `pm_debug_messages`, and a loosened journald sync interval — live in `hosts/thinkpad-e14-gen5/quirks.sh` and are checked separately by `/host-check`. This skill's only job is: did the last boot end badly, and if so, hand over what was already captured. Follow-up on any specific signature is case-by-case and out of scope here — see the relevant `incidents/` entry or open a new one. ## When this runs Invoked from the `handover` skill's read-mode, after that skill's own present/absent branch on `.claude/handover.md` — both branches, since an unclean shutdown isn't tied to whether a handover file exists. Not a standalone entry point; don't invoke this before `handover` has run. ## Detection ``` last -x | awk ' /^reboot/ { n++; if (n==2) exit } /crash/ { print; exit } ' ``` `last -x` reads wtmp and explicitly marks a **login session line** `crash` instead of a clean end-time range when no matching "shutdown system down" record exists for it — i.e. the machine came back up without ever writing a clean shutdown record. The `reboot` lines themselves do *not* carry this tag on this system: a `reboot` entry just perpetually reads `still running` once no shutdown record follows it, which is true both for the genuinely current boot and for a crashed prior boot that never got a matchi