Start Monitor Workers Loop

Run /monitor-workers in a continuous loop. This is the default command for the manager session in a parallel agent swarm.

Usage

# Start with default 15-minute interval
/monitor-loop

# Custom interval
/monitor-loop 5

What This Does

Schedules /autocoder:monitor-workers to run on a recurring interval using the /loop command. Each iteration:

1. Checks worker status via cmux/tmux
2. Detects stale "working" labels
3. Dispatches idle workers to unblocked issues
4. Runs /autocoder:review-blocked when all issues are blocked
5. Deploys when all work is complete

How It Works

Mode 1: /loop command (preferred, when CronCreate tool is available)

Uses the native CronCreate mechanism to schedule recurring /autocoder:monitor-workers calls. No stop hook or settings.json modification needed.

Mode 2: Manual loop (fallback for older Claude Code versions)

Runs /autocoder:monitor-workers in a bash sleep loop, polling at the configured interval.

Instructions

# Parse arguments
INTERVAL_MINUTES="${1:-15}"

echo ""
echo "🔄 Starting monitor-workers loop"
echo "   Interval: ${INTERVAL_MINUTES}m"
echo ""
echo "This will:"
echo "  • Check worker status every ${INTERVAL_MINUTES} minutes"
echo "  • Dispatch idle workers to unblocked issues"
echo "  • Run /review-blocked when all issues need human review"
echo "  • Deploy to ey-staging when all work completes"
echo ""

Detect /loop availability and choose mode

Check whether the CronCreate tool is available (it appears in the available deferred tools list when Claude Code supports the /loop command).

If CronCreate IS available → Use /loop mode (preferred):

echo "✅ Using /loop command (CronCreate available)"
echo "To stop: use CronDelete to remove the scheduled job"
echo ""

Run the first iteration immediately, then schedule the loop:

First, invoke the monitor-workers skill directly (using the Skill tool) to do an immediate check. Then use the loop skill to schedule recurring checks:

Use the Skill tool to invoke: monitor-workers

After the first iteration completes:

Use the Skill tool to invoke: loop
With args: ${INTERVAL_MINUTES}m /autocoder:monitor-workers

This schedules /autocoder:monitor-workers to run every INTERVAL_MINUTES minutes using CronCreate.

---

If CronCreate is NOT available → Use manual loop (fallback):

echo "⚠️  /loop command not available — using manual sleep loop"
echo "To stop: Ctrl+C"
echo ""

Run the first iteration immediately, then loop with sleep:

Use the Skill tool to invoke: monitor-workers

After each iteration, sleep and repeat:

INTERVAL_SECONDS=$((INTERVAL_MINUTES * 60))
echo ""
echo "💤 Next check in ${INTERVAL_MINUTES} minutes..."
sleep "$INTERVAL_SECONDS"

Then invoke monitor-workers again via the Skill tool. Repeat until manually stopped (Ctrl+C) or all work is complete and deployed.