Skill מוניטורינג של אוטומציות ל-Claude

מתי להשתמש

"Monitoring", "Alerts", "Workflow uptime", "Health check", "SLO/SLA tracking".

הוראות עבודה

1. What to Monitor

Health

• Success rate — % of successful runs.
• Error rate — % failed.
• Latency — p50, p95, p99.
• Volume — runs per period.

Cost

• Operations/Tasks consumed.
• AI tokens used.
• Per-workflow cost.

Business Impact

• Records processed.
• Tasks completed (e.g., leads routed).

2. Alerting Rules

Good Alerts (Actionable)

• ✅ "Workflow X failed 5+ times in 10 min".
• ✅ "Cost > $X this week (vs $Y last)".
• ✅ "Latency p95 > 30 seconds".

Bad Alerts (Noise)

• ❌ "Workflow ran successfully" (silence is golden).
• ❌ Every single error (alert fatigue).
• ❌ "Health check passed".

3. Tools

Built-in

• Zapier History + Slack integration.
• Make.com — execution log.
• n8n — execution history.

External

• Better Uptime — endpoint monitoring.
• Datadog — comprehensive APM.
• Sentry — error tracking.
• PagerDuty — incident management.

4. Sample Monitoring Workflow

Setup: Zapier scheduled task (hourly)
1. Pull execution logs (Zapier API)
2. Calculate:
   - Success rate
   - Error count
   - Avg latency
3. If error rate > 5% → Slack #alerts
4. If latency > threshold → Slack #alerts
5. Daily summary at 9 AM → email to ops lead

5. SLO (Service Level Objective)

Define

• Critical workflow: 99.9% uptime, p95 < 5 sec.
• Standard: 99% uptime, p95 < 30 sec.
• Best-effort: 95% uptime, no latency SLO.

Track

• Weekly review.
• Quarterly: adjust SLOs based on data.

6. Israel Specifics

• Israeli timezone (UTC+2/+3) in alerts.
• WhatsApp alerts common.
• On-call rotation for critical workflows.

7. אסיים בהמלצה.

קלט נדרש

פלט צפוי

דגלים אדומים

• 🚨 No monitoring — silent failures.
• 🚨 Alert on every error — fatigue.
• 🚨 No on-call for critical workflows.

פרומפט לדוגמה

Critical workflow ב-Make. Monitoring + alerting plan.

Cost ב-Zapier מתפוצץ. איך לעקוב?

SLO ל-lead routing workflow.

© 2026 Automation Expert Pro | גרסה 1.0.0