[JOsacky]

Why

Skillhook Cloud cannot tell why a first-value machine went offline. A closed terminal (foreground `skillhook serve`), a Linux logout with lingering off, and a deliberate stop all show up as the same `link.stopped`. This blocks attribution for two cases: the only external Linux machine (2c2b1258, stopped with SIGTERM twice even though the unit has `Restart=always`) and org aee7a873's macOS machine, which stopped 6 minutes after its first useful job. Ops backlog items: MI-020, MI-029 and PC-010.

What

• `service.ts`: pure `parseLinger` and `runModeOf`, plus `lingerStatus()`, which runs `loginctl show-user --property=Linger --value`. The call is read-only and never uses sudo; any failure gives `unknown`.
• `health.ts` service check: the details now carry `run_mode` (`service` when the answering server's pid is the launchd/systemd service pid, `foreground` otherwise, `unknown` with no server) and, for an installed systemd unit, `linger`.
  • A running systemd service with `linger: no` is now a `warn`, with the hint `loginctl enable-linger $USER`.
  • Under `serve`, the server's own pid is used, so the cloud health report carries these fields. No cloud or protocol change is needed: check details are loose records.
• `service install` on Linux warns when lingering is off and skips the tip when it is on.
• Docs row in `docs/operations.md` and a CHANGELOG entry.

Privacy

• Only enum values are sent: no username, no paths.

Tests

• New unit tests for `parseLinger` and `runModeOf`.
• `tsc` is clean. `service.test.ts` and `health.test.ts` pass locally (20 of 20).
• The first local health-test run timed out under a load average of about 465–700 on the operator Mac. The rerun passed, and CI runs the full suite.

Validation after release

• On the operator Mac, the health tab shows `run_mode: service`.
• A `skillhook serve` started in a terminal shows `foreground`.
• The first Linux report shows `linger`.

🤖 Generated with Claude Code