| name | progress-heartbeat |
| description | Full sweep of all interns. Only use when explicitly asked for a heartbeat, pulse check, or sweep. |
Progress Heartbeat
Full sweep of every intern. Query them all, assess each one, check disk usage, take action, update tasks, deliver a consolidated report.
When to Use
- User says "heartbeat", "pulse check", "how's everyone doing?"
- Routine sweep to check nothing has stalled
- After delegating a batch of work
Procedure
1. Discover All Interns
intern_list
Also read tasks.md to cross-reference. If there are no interns, tell the user and stop.
2. Check Disk Usage
du -sh ./interns/*/
Flag anything over 20MB. For big ones, dig in with du -sh ./interns/<name>/*.
3. Query All Interns
For each live intern, use intern_status to query them. You can call multiple intern_status tools in parallel:
intern_status name: "<name>" question: "Heartbeat. Be brief: 1. Original task (one line) 2. What you've completed 3. What's remaining 4. Blocked? (yes/no, why) 5. Confidence to finish (high/medium/low)"
For busy interns, intern_status returns their latest output without interrupting them.
4. Assess and Act
| Category | Criteria | Action |
|---|
| Done | Claims complete, credible | Verify artifacts, mark done |
| On track | Clear progress, confident | Log it, move on |
| Slow | Some progress, less than expected | Nudge via intern_send |
| Stuck | Blocked, confused, no progress | Firm redirect via intern_steer or intern_send |
| Lost | Incoherent, looping, off-task | intern_kill, clean up |
5. Update Task Tracker
Single comprehensive update to tasks.md — all interns, disk notes, actions taken.
6. Report to User
Table format:
## Heartbeat Report — YYYY-MM-DD HH:MM
| Intern | Status | Progress | Disk | Action |
|--------|--------|----------|------|--------|
| name | icon | summary | size | what you did |
### Summary
- X interns checked, Y done, Z nudged, ...
- Total disk: ...
### Needs Attention
- anything requiring user decision
Notes
- Gather all statuses first, then act — full picture before taking action
- Disk growth is a signal — runaway loops, unnecessary downloads, verbose logging
- If multiple interns are stuck on related things, tell the user — systemic issue
- Don't nudge completed interns
