v0.6.1

Restart-engine hardening from a re-audit of the shutdown/restart path:

• Retries stop when they cannot win: an old instance still holding the single-instance lock is reported as lock-held-by-other, and a launch path that reported no pid at all is declared dead after ~6s instead of consuming the whole start budget.
• The helper watches the new instance after saying OK (outside the lock): died / teardown / replaced are recorded instead of leaving a false success that never changes.
• Same-named processes younger than the kill are stopped before the relaunch, so a shell respawning its host cannot race the new instance for the port.
• Hand-run mode (-Port 0) no longer polls a budget it cannot satisfy, and no longer reports a healthy relaunch as teardown with exit code 3.
• WMI calls carry a 20s operation timeout: a hung CIM call falls through to the explorer fallback instead of leaving the app stopped with nobody left to relaunch it.
• The status record carries servingPid (the host pid the port actually named), not just the pid a launch path reported.
• New testbed/ end-to-end tests against a fake app (scenarios A/B/C plus hand-run mode); docs corrected (readiness is the plugin's own status route, because the core /health route is behind the web token).

Install: dsh plugin --profile desktop add github:nanshang328/dsh-power

v0.6.0

One draggable power card (left = shut down, right = restart, two-click safety).

• CSRF-hardened endpoints (x-dsh-power header)
• Multi-instance-safe restart (kills only this instance's process tree)
• Foreground bring-up after restart
• schtasks cleanup, confirmPower/confirmMs config keys

Install: dsh plugin --profile desktop add github:nanshang328/dsh-power