Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Removes the egregore watchdog daemon and its associated files. Use when stopping automated session relaunching or cleaning up egregore infrastructure.
.claude/skills/athola-uninstall-watchdog/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -47% | 0% |
Removes the egregore watchdog daemon and cleans up all files created by the install-watchdog skill. After uninstalling, egregore sessions will no longer be relaunched automatically.
the verify commands below).
bashOS=$(uname -s)
macOS (launchd):
bashPLIST=~/Library/LaunchAgents/com.egregore.watchdog.plist # Unload the agent (stops it if running) launchctl unload "$PLIST" 2>/dev/null # Remove the plist file rm -f "$PLIST"
Linux (systemd):
bash# Stop and disable the timer and service systemctl --user stop egregore-watchdog.timer 2>/dev/null systemctl --user disable egregore-watchdog.timer 2>/dev/null # Remove unit files rm -f ~/.config/systemd/user/egregore-watchdog.timer rm -f ~/.config/systemd/user/egregore-watchdog.service # Reload systemd to pick up the removal systemctl --user daemon-reload
bash# Remove pidfile if present rm -f ~/.egregore/watchdog.pid # Remove watchdog log rm -f ~/.egregore/watchdog.log
macOS:
bashlaunchctl list | grep egregore # Should produce no output
Linux:
bashsystemctl --user list-timers | grep egregore # Should produce no output
Report to the user that the watchdog has been removed and automatic relaunching is disabled.
| File | Platform | Purpose | |------|----------|---------| | ~/Library/LaunchAgents/com.egregore.watchdog.plist | macOS | launchd agent definition | | ~/.config/systemd/user/egregore-watchdog.timer | Linux | systemd timer unit | | ~/.config/systemd/user/egregore-watchdog.service | Linux | systemd service unit | | ~/.egregore/watchdog.pid | both | PID of last watchdog run | | ~/.egregore/watchdog.log | macOS | watchdog output log |
~/Library/LaunchAgents/com.egregore.watchdog.plistdoes not exist after uninstall completes
~/.config/systemd/user/egregore-watchdog.timer andegregore-watchdog.service are both removed and systemctl --user daemon-reload has run
grep egregore (macOS) or systemctl --user list-timers | grep egregore (Linux)
~/.egregore/watchdog.pid and ~/.egregore/watchdog.log areremoved (or absent if they were never created)
automatic relaunching is disabled
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,246 | 3,327 | -70% | 1 | 1 | 0% | 1,795 | 1,270 | -29% | 0 | 0 | — |
case-02 | fail→fail | 8,952 | 9,019 | +1% | 1 | 1 | 0% | 1,514 | 1,428 | -6% | 0 | 0 | — |
case-03 | fail→fail | 7,167 | 8,784 | +23% | 1 | 1 | 0% | 1,348 | 1,363 | +1% | 0 | 0 | — |
case-04 | pass→pass | 23,748 | 6,363 | -73% | 1 | 1 | 0% | 2,142 | 1,940 | -9% | 0 | 0 | — |
case-05 | pass→pass | 11,208 | 5,569 | -50% | 1 | 1 | 0% | 1,930 | 1,861 | -4% | 0 | 0 | — |
case-06 | pass→pass | 11,880 | 4,725 | -60% | 1 | 1 | 0% | 2,044 | 1,546 | -24% | 0 | 0 | — |
case-07 | fail→pass | 11,327 | 3,414 | -70% | 1 | 1 | 0% | 1,868 | 1,418 | -24% | 0 | 0 | — |
case-08 | pass→pass | 9,289 | 4,028 | -57% | 1 | 1 | 0% | 1,526 | 1,487 | -3% | 0 | 0 | — |
case-09 | fail→pass | 11,571 | 3,745 | -68% | 1 | 1 | 0% | 2,023 | 1,505 | -26% | 0 | 0 | — |
case-10 | fail→pass | 11,709 | 4,302 | -63% | 1 | 1 | 0% | 2,050 | 1,560 | -24% | 0 | 0 | — |
case-11 | pass→pass | 2,909 | 2,322 | -20% | 1 | 1 | 0% | 416 | 1,178 | +183% | 0 | 0 | — |
case-12 | fail→pass | 13,187 | 2,189 | -83% | 1 | 1 | 0% | 2,063 | 1,121 | -46% | 0 | 0 | — |
case-13 | fail→pass | 11,967 | 1,622 | -86% | 1 | 1 | 0% | 1,894 | 1,011 | -47% | 0 | 0 | — |
case-14 | pass→pass | 6,462 | 1,708 | -74% | 1 | 1 | 0% | 1,097 | 1,069 | -3% | 0 | 0 | — |
case-15 | fail→pass | 6,505 | 2,034 | -69% | 1 | 1 | 0% | 1,036 | 1,122 | +8% | 0 | 0 | — |
case-16 | pass→pass | 17,724 | 10,961 | -38% | 1 | 1 | 0% | 3,099 | 2,799 | -10% | 0 | 0 | — |
case-17 | fail→pass | 8,876 | 2,274 | -74% | 1 | 1 | 0% | 1,388 | 1,186 | -15% | 0 | 0 | — |
case-18 | fail→pass | 19,634 | 2,207 | -89% | 1 | 1 | 0% | 1,671 | 1,173 | -30% | 0 | 0 | — |
case-19 | pass→pass | 6,632 | 4,422 | -33% | 1 | 1 | 0% | 1,146 | 1,591 | +39% | 0 | 0 | — |
case-20 | fail→pass | 15,610 | 5,093 | -67% | 1 | 1 | 0% | 2,742 | 1,697 | -38% | 0 | 0 | — |
case-21 | pass→pass | 6,272 | 2,586 | -59% | 1 | 1 | 0% | 1,200 | 1,272 | +6% | 0 | 0 | — |
case-22 | fail→pass | 20,715 | 2,213 | -89% | 1 | 1 | 0% | 920 | 1,228 | +33% | 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 20 counted toward the lift figure. The other 2 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 +45 percentage points is the difference between those two pass rates over the 20 comparable cases. 2 cases got worse with the skill loaded, and they are 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.