Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Installs egregore watchdog daemon via launchd or systemd for autonomous relaunching. Use when setting up egregore on a new machine. Do not use on CI/CD runners.
.claude/skills/athola-install-watchdog/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -46% | 0% |
Installs a persistent daemon that monitors the egregore process and relaunches it when needed. The daemon uses the OS-native scheduler (launchd on macOS, systemd on Linux) to run a watchdog script every 5 minutes.
The watchdog script checks:
If all conditions are met, it launches a new egregore session.
egregore init)when you want autonomous relaunching.
bashOS=$(uname -s)
macOS (launchd):
bashbash plugins/egregore/scripts/install_launchd.sh
This script creates a plist at ~/Library/LaunchAgents/com.egregore.watchdog.plist that runs the watchdog script every 300 seconds (5 minutes).
Linux (systemd):
bashbash plugins/egregore/scripts/install_systemd.sh
This script creates a systemd timer and service unit at ~/.config/systemd/user/ that fires every 5 minutes.
macOS:
bashlaunchctl list | grep egregore
Expected output: a line containing com.egregore.watchdog with a PID or - status.
Linux:
bashsystemctl --user status egregore-watchdog.timer
Expected output: active (waiting) status.
Report the installation result, the schedule interval, and the log file location:
~/.egregore/watchdog.logjournalctl --user -u egregore-watchdogTo remove the watchdog, run:
Skill(egregore:uninstall-watchdog)Or invoke the uninstall skill directly via the command /egregore:uninstall-watchdog.
as the current user, not root.
loginctl enable-linger $USER to enable user services without an active login session.
the scheduler is loaded (launchctl list or systemctl --user list-timers).
~/Library/LaunchAgents/com.egregore.watchdog.plistexists after the install script completes
~/.config/systemd/user/egregore-watchdog.timerexists after the install script completes
grep egregore returns a line (macOS), or systemctl --user status egregore-watchdog.timer shows active (waiting) (Linux)
reported to the user after installation
permission denied occurs, the error is reported with theremediation step (run as current user, not root)
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 10,821 | 5,100 | -53% | 1 | 1 | 0% | 1,920 | 1,203 | -37% | 0 | 0 | — |
case-02 | fail→fail | 14,684 | 5,542 | -62% | 1 | 1 | 0% | 741 | 1,135 | +53% | 0 | 0 | — |
case-03 | fail→fail | 5,245 | 6,315 | +20% | 1 | 1 | 0% | 240 | 1,235 | +415% | 0 | 0 | — |
case-04 | fail→pass | 16,414 | 3,579 | -78% | 1 | 1 | 0% | 1,623 | 1,342 | -17% | 0 | 0 | — |
case-05 | pass→pass | 13,228 | 4,202 | -68% | 1 | 1 | 0% | 2,032 | 1,547 | -24% | 0 | 0 | — |
case-06 | fail→pass | 12,565 | 4,628 | -63% | 1 | 1 | 0% | 2,174 | 1,683 | -23% | 0 | 0 | — |
case-07 | fail→pass | 8,828 | 2,804 | -68% | 1 | 1 | 0% | 1,398 | 1,229 | -12% | 0 | 0 | — |
case-08 | fail→pass | 6,815 | 2,640 | -61% | 1 | 1 | 0% | 1,154 | 1,238 | +7% | 0 | 0 | — |
case-09 | fail→pass | 11,564 | 2,030 | -82% | 1 | 1 | 0% | 2,050 | 1,117 | -46% | 0 | 0 | — |
case-10 | fail→pass | 5,633 | 1,936 | -66% | 1 | 1 | 0% | 972 | 1,106 | +14% | 0 | 0 | — |
case-11 | fail→pass | 10,480 | 2,037 | -81% | 1 | 1 | 0% | 1,694 | 1,113 | -34% | 0 | 0 | — |
case-12 | fail→pass | 9,232 | 3,279 | -64% | 1 | 1 | 0% | 1,611 | 1,391 | -14% | 0 | 0 | — |
case-13 | pass→pass | 12,220 | 5,425 | -56% | 1 | 1 | 0% | 2,223 | 1,692 | -24% | 0 | 0 | — |
case-14 | pass→pass | 8,533 | 3,488 | -59% | 1 | 1 | 0% | 1,441 | 1,378 | -4% | 0 | 0 | — |
case-15 | pass→pass | 11,107 | 6,297 | -43% | 1 | 1 | 0% | 2,065 | 1,962 | -5% | 0 | 0 | — |
case-16 | pass→pass | 5,568 | 1,643 | -70% | 1 | 1 | 0% | 976 | 1,008 | +3% | 0 | 0 | — |
case-17 | fail→pass | 10,532 | 2,283 | -78% | 1 | 1 | 0% | 1,715 | 1,210 | -29% | 0 | 0 | — |
case-18 | fail→pass | 9,726 | 3,661 | -62% | 1 | 1 | 0% | 1,646 | 1,445 | -12% | 0 | 0 | — |
case-19 | fail→pass | 9,063 | 3,836 | -58% | 1 | 1 | 0% | 1,625 | 1,468 | -10% | 0 | 0 | — |
case-20 | fail→pass | 12,031 | 1,376 | -89% | 1 | 1 | 0% | 2,022 | 973 | -52% | 0 | 0 | — |
case-21 | fail→pass | 4,985 | 1,682 | -66% | 1 | 1 | 0% | 883 | 1,048 | +19% | 0 | 0 | — |
case-22 | fail→pass | 10,494 | 4,613 | -56% | 1 | 1 | 0% | 1,602 | 1,538 | -4% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +64 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.