| name | monitor-loop |
| description | Actively monitor a running compound-agent infinity loop, checking progress every N minutes (default 5). Reports epic status, current activity, and flags issues. |
Monitor Loop
Actively monitors a running ca loop infinity loop session, checking back at regular intervals to report progress and flag issues.
Usage
/monitor-loop # Monitor every 5 minutes (default)
/monitor-loop 2 # Monitor every 2 minutes
/monitor-loop 10 # Monitor every 10 minutes
How It Works
On each check cycle, run ALL of these commands and synthesize a status report:
1. Verify the loop is running
screen -ls 2>/dev/null | grep beads-loop || echo "NO SCREEN SESSION"
If no session found, report LOOP DOWN and stop monitoring.
2. Check loop output for latest events
tail -20 .compound-agent/agent_logs/loop-output*.log 2>/dev/null | tail -20
Look for: epic completions, review phases, failures, skips, crashes.
3. Check current epic progress
bd list --status=in_progress 2>/dev/null
bd stats 2>/dev/null
If bd is blocked (Dolt lock), note it and check process list instead.
4. Check agent activity
tail -c 500 .compound-agent/agent_logs/loop_beads-gui-*-$(date +%Y-%m-%d)*.log 2>/dev/null | tail -10
5. Check for stuck processes
ps aux | grep 'playwright' | grep -v grep | wc -l
ps aux | grep -E 'vite|dolt sql-server' | grep -v grep | wc -l
Status Report Format
Each check should produce a concise status like:
**Status: [Epic Name] -- [phase] ([N/M] done)**
- Screen: running/down
- Current: [what the agent is doing]
- Closed: X / Y total
- Issues: [any problems detected]
When to Flag Issues
- LOOP DOWN: Screen session gone -- needs relaunch
- STUCK: Agent log hasn't updated in >10 minutes and no processes running
- DOLT LOCK:
bd commands failing -- stale dolt sql-server process
- TEST FAILURES: Agent iterating on test fixes for >3 cycles
- MEMORY: System memory below 90% free
Scheduling
After each check, use ScheduleWakeup to schedule the next one:
ScheduleWakeup({
delaySeconds: interval * 60, // Convert minutes to seconds
reason: "Check loop -- [brief status]",
prompt: "/monitor-loop N"
})
Use delaySeconds of 270 for 5-min intervals (stays within prompt cache window).
Important Notes
- Parse the interval from the user's args (default: 5 minutes)
- Keep reports concise -- the user doesn't need every log line
- If the loop finishes (all epics done), report completion and stop monitoring
- If
bd is blocked by Dolt lock, check processes instead -- don't keep retrying bd
