Automation Workflows

Hermes Agent’s cron system and tool integration make it perfect for building automated workflows. This guide covers common automation patterns and best practices.

Common Automation Patterns

1. Daily Morning Briefing

# Create a daily briefing job
hermes cron create "every day at 8am" \
  "Check my calendar, summarize unread emails, and check Hacker News top stories. Send a briefing to Telegram." \
  --name "morning-briefing" \
  --deliver telegram

2. Server Health Monitoring

# Create a watchdog script
cat > ~/.hermes/scripts/health-check.sh << 'EOF'
#!/bin/bash
CPU=$(top -bn1 | grep "Cpu(s)" | awk '{print $2}' | cut -d'%' -f1)
MEM=$(free | grep Mem | awk '{printf "%.0f", $3/$2 * 100.0}')
if (( $(echo "$CPU > 90" | bc -l) )) || (( $MEM > 90 )); then
  echo "⚠️ Alert: CPU ${CPU}% | Memory ${MEM}%"
  exit 1
fi
EOF

chmod +x ~/.hermes/scripts/health-check.sh

# Schedule the check
hermes cron create "every 5m" \
  --no-agent \
  --script health-check.sh \
  --deliver telegram \
  --name "server-health"

3. Automated Backups

# Backup job with agent reasoning
cronjob(
  action="create",
  schedule="0 2 * * *",  # 2 AM daily
  name="daily-backup",
  prompt="""
  1. Check disk space in /backup
  2. Create a compressed backup of ~/projects
  3. Verify backup integrity
  4. Remove backups older than 7 days
  5. Report backup size and status
  """,
  deliver="telegram"
)

4. Content Monitoring

# Monitor a website for changes
cronjob(
  action="create",
  schedule="every 1h",
  name="price-monitor",
  prompt="""
  Check https://example.com/product and extract the current price.
  Compare with the previous price stored in ~/.hermes/data/prices.json.
  If price dropped by more than 10%, notify with [SILENT] removed.
  Otherwise respond with [SILENT].
  """,
  deliver="telegram"
)

Workflow Composition

Chaining Jobs with Context

# Job 1: Collect data
cronjob(
  action="create",
  schedule="0 7 * * *",
  name="collect-news",
  prompt="Fetch top AI/ML stories from Hacker News and save to ~/.hermes/data/news.md"
)

# Job 2: Process data (depends on Job 1)
cronjob(
  action="create",
  schedule="30 7 * * *",
  name="summarize-news",
  context_from="<job1_id>",
  prompt="Read ~/.hermes/data/news.md and create a 3-bullet summary"
)

# Job 3: Deliver (depends on Job 2)
cronjob(
  action="create",
  schedule="0 8 * * *",
  name="deliver-news",
  context_from="<job2_id>",
  prompt="Format the summary for Telegram and deliver"
)

Using Skills in Automation

# Create a bundle for data processing
hermes bundles create data-pipeline \
  --skill web-scraper \
  --skill data-cleaner \
  --skill report-generator

# Use the bundle in a cron job
hermes cron create "every 6h" \
  --skill data-pipeline \
  "Process new sales data and generate a report" \
  --name "sales-report"

Best Practices

1. Error Handling

Always include error handling in your automation prompts:

Check the API status. If it fails:
1. Wait 30 seconds and retry up to 3 times
2. If still failing, report the error but don't spam
3. Use [SILENT] for expected transient failures

2. Rate Limiting

Be respectful of external APIs:

When calling external APIs:
- Add delays between requests (sleep 1-2 seconds)
- Respect rate limits (check response headers)
- Cache results when appropriate
- Use exponential backoff for retries

3. Resource Management

For resource-intensive tasks:
- Check available disk space before large operations
- Monitor memory usage during processing
- Clean up temporary files after completion
- Use --no-agent mode for simple scripts

4. Security

For automated tasks:
- Store credentials in ~/.hermes/.env, never in prompts
- Use read-only operations when possible
- Validate inputs before processing
- Log sensitive operations for audit

Example: Complete CI/CD Pipeline

# Job 1: Monitor repository
cronjob(
  action="create",
  schedule="every 15m",
  name="repo-monitor",
  prompt="""
  Check GitHub repo for new commits on main branch.
  If new commits found:
  1. Pull changes
  2. Run tests (pytest)
  3. If tests pass, build Docker image
  4. Report status
  """,
  deliver="slack"
)

Troubleshooting Automation

Jobs Not Running

# Check gateway status
hermes gateway status
hermes cron list

# View logs
hermes logs --follow

# Manually trigger a job
hermes cron run <job_id>

Delivery Failures

# Check delivery target configuration
hermes config show

# Verify platform credentials
cat ~/.hermes/.env | grep -E "(TELEGRAM|DISCORD|SLACK)"

# Test delivery manually
hermes chat -q "test message" --deliver telegram

Learn More

• Cron Jobs - Complete cron documentation
• Skills System - Create reusable automation components
• Sub-Agents - Parallel task execution