Heartbeat Manager

Heartbeat Manager

Important: Instruction Template This skill provides a checklist and configuration format for Claude Code to follow. There is no automatic scheduler - heartbeat checks run only when Claude reads and follows this template during a session. For true automation, integrate with external schedulers (cron, systemd timers) that invoke Claude Code periodically.

Goal

Enable proactive agent behavior through scheduled heartbeat checks that run without user prompting.

Concept

A heartbeat is a periodic check that allows the agent to:

• Monitor for items requiring attention
• Take autonomous action on routine tasks
• Alert the user only when needed
• Maintain awareness of ongoing processes

Key Principle: Silence = all systems normal. Only notify when action is required.

Heartbeat Configuration

Location

automation/HEARTBEAT.md or .agent/HEARTBEAT.md

Format

# Heartbeat Configuration

## Schedule
- **Interval:** 30m
- **Active Hours:** 08:00 - 22:00
- **Timezone:** America/New_York

## Checks

### Communications
- [ ] Check email for urgent messages
- [ ] Review Slack/Discord for direct mentions
- [ ] Check GitHub notifications

### Monitoring
- [ ] Verify background jobs completed
- [ ] Check for CI/CD failures
- [ ] Monitor system alerts

### Proactive Tasks
- [ ] If idle > 2 hours with pending tasks, remind
- [ ] If PR approved, suggest merge
- [ ] If deadline approaching, alert

## Notification Rules
- **Urgent:** Notify immediately
- **Normal:** Batch and notify at next check
- **Low:** Log only, no notification

Process

Creating a Heartbeat

1. Identify Monitoring Needs

   • What should be checked periodically?
   • What requires user attention vs autonomous action?
   • What's the appropriate frequency?

2. Define Checks For each check, specify:

   • What to check
   • How to determine if action needed
   • What action to take (notify/act/log)

3. Set Schedule

   • Interval: How often to run (15m, 30m, 1h)
   • Active hours: When to run (respect user's schedule)
   • Timezone: For consistent timing

4. Configure Notifications

   • Where to deliver (chat, email, log)
   • Urgency levels and routing
   • Batching rules

Running a Heartbeat

When heartbeat triggers:

1. Load Configuration Read HEARTBEAT.md for current checks

2. Execute Checks Run each check in the list:

   • Evaluate condition
   • Determine if action needed
   • Log result

3. Process Results

   • Nothing to report: Silent, log only
   • Items found: Compile notification
   • Urgent items: Immediate alert

4. Deliver Notifications If items require attention:

   🔔 Heartbeat Check (HH:MM)
   
   ## Requires Attention
   - [Item 1]
   - [Item 2]
   
   ## Completed Autonomously
   - [Action taken]
   

5. Update Memory Log heartbeat results to daily notes

Example Configurations

Developer Heartbeat

# Developer Heartbeat

## Schedule
- **Interval:** 30m
- **Active Hours:** 09:00 - 18:00

## Checks

### Code Review
- [ ] PRs awaiting my review
- [ ] My PRs with new comments
- [ ] CI failures on my branches

### Tasks
- [ ] Blocked tasks that may be unblocked
- [ ] Approaching deadlines (< 24h)

### Communications
- [ ] Slack mentions in engineering channels
- [ ] GitHub @mentions

## Auto-Actions
- Mark stale PRs for follow-up
- Draft responses to simple questions

Personal Assistant Heartbeat

# Personal Assistant Heartbeat

## Schedule
- **Interval:** 1h
- **Active Hours:** 07:00 - 22:00

## Checks

### Calendar
- [ ] Upcoming meetings (next 2 hours)
- [ ] Meetings requiring prep
- [ ] Scheduling conflicts

### Email
- [ ] Urgent emails (VIP senders)
- [ ] Emails awaiting response > 24h

### Tasks
- [ ] Overdue items
- [ ] Items due today

## Auto-Actions
- Prepare meeting briefs
- Draft email responses for review

Heartbeat vs Cron

Use heartbeat for: "Check if X needs attention" Use cron for: "Do Y at specific time"

Commands

• "Configure heartbeat" - Set up or modify heartbeat
• "Run heartbeat now" - Execute immediate check
• "Pause heartbeat" - Temporarily disable
• "Show heartbeat status" - Display configuration and last run

Integration

With Memory Manager

• Log heartbeat results to daily notes
• Track patterns over time in MEMORY.md

With Task Processor

• Check for tasks that can be progressed
• Alert on blocked or stale tasks

With Communication Skills

• Check relevant channels
• Draft responses when appropriate

Autonomous Action Guidelines

May Act Autonomously

• Checking status of systems
• Logging information
• Preparing summaries
• Marking items for attention

Must Ask First

• Sending external messages
• Modifying data
• Taking irreversible actions
• Spending money

References

• See AUTONOMOUS_BOOTUP_SPEC.md for architecture
• See memory-manager for logging patterns
• See cron documentation for scheduled tasks