daemon-watchdog

name	daemon-watchdog
description	Check all host-side daemons are running and healthy. Auto-restart what can be restarted, alert on what can't. Run on heartbeat or /daemon-status.
version	1.0.0
author	ops-officer
tags	["daemons","health","watchdog","infrastructure","tap","backbone","executor"]
trigger	{"command":"/daemon-status","keywords":["daemon health","daemon status","process check","tap status","backbone status","executor status"]}

daemon-watchdog

Check all host-side daemons and report status. Auto-restart recoverable processes.

Monitored daemons

Daemon	Process match	Auto-restart	Owner
Tap daemon	python3 tap-daemon.py	Yes	Ops Officer
host-ops-executor	host-ops-executor	No (systemd)	Ops Officer
backbone-listener	backbone-listener.py	Yes	Ops Officer
relay-handoff-watcher	relay-handoff-watcher.py	Yes	Ops Officer
telegram-listener	telegram-listener.py	Yes (cron)	Ops Officer
Bridge dashboard (main)	bridge/dashboard-api.py	Alert only	Ops Officer
Bridge dashboard (dev)	bridge-dev/dashboard-api.py	Alert only	Ops Officer
Bridge dashboard (corinne)	bridge-corinne/dashboard-api.py	Alert only	Ops Officer
Ollama	ollama serve	Alert only	Ops Officer
Gateway container	openclaw-gateway	Yes (stability-monitor)	Ops Officer

Procedure

1. Run the check:

   echo "=== Daemon Watchdog ===" && \
   for proc in "python3 tap-daemon.py:Tap" "host-ops-executor:Executor" "backbone-listener.py:Backbone" "relay-handoff-watcher.py:Handoff Watcher" "telegram-listener.py:Telegram Listener" "ollama serve:Ollama"; do
     NAME="${proc#*:}"; PATTERN="${proc%%:*}"
     if ps aux | grep "[$( echo "$PATTERN" | cut -c1)]$(echo "$PATTERN" | cut -c2-)" | grep -qv grep 2>/dev/null; then
       echo "  ✅ $NAME"
     else
       echo "  ❌ $NAME — DOWN"
     fi
   done && \
   for BRIDGE in bridge bridge-dev bridge-corinne; do
     if ps aux | grep "[d]ashboard-api.py" | grep -q "$BRIDGE"; then
       echo "  ✅ Bridge ($BRIDGE)"
     else
       echo "  ❌ Bridge ($BRIDGE) — DOWN"
     fi
   done
   

2. For machine-readable output (agent-compatible):

On heartbeat

Run the quick check. If any daemon is down:

1. Attempt auto-restart for recoverable daemons
2. If restart fails, create an ops.db task with urgency=critical
3. Alert via Discord ops channel (never Telegram — daemon failures are routine ops)

Escalation

• 1 restart attempt per 5-minute cycle (stability-monitor handles this)
• If same daemon fails 3 times in 15 minutes, escalate to Robert via Telegram
• Chart the failure pattern

Important

• Use ps aux | grep "[f]irst-char-bracket-trick" pattern — NOT pgrep -f
• pgrep -f has false positives when Claude Code sessions are running (CC bash wrappers contain process names as strings)
• stability-monitor.sh is the automated enforcement; this skill is for on-demand checks and agent awareness