observability-grafana

name	observability-grafana
description	Grafana dashboard and metrics analysis. Use when querying dashboards, panels, Prometheus metrics via Grafana, checking datasources, reviewing alerts, or creating dashboards from templates.
allowed-tools	Bash(python *)

Grafana Dashboard Analysis

Authentication

IMPORTANT: Credentials are injected automatically by a proxy layer. Do NOT check for GRAFANA_API_KEY in environment variables - it won't be visible to you. Just run the scripts directly; authentication is handled transparently.

Configuration environment variables you CAN check (non-secret):

• GRAFANA_URL - Grafana instance URL (e.g., https://grafana.company.com)

---

Available Scripts

All scripts are in .claude/skills/observability-grafana/scripts/

list_dashboards.py - Search/List Dashboards

python .claude/skills/observability-grafana/scripts/list_dashboards.py [--query "kubernetes"]

get_dashboard.py - Get Dashboard Details

python .claude/skills/observability-grafana/scripts/get_dashboard.py --uid DASHBOARD_UID

query_prometheus.py - Query Prometheus via Grafana

python .claude/skills/observability-grafana/scripts/query_prometheus.py --query "PROMQL" [--time-range 60] [--step 1m] [--datasource-id ID]

# Examples (auto-detects Prometheus datasource):
python .claude/skills/observability-grafana/scripts/query_prometheus.py --query "rate(http_requests_total[5m])"
python .claude/skills/observability-grafana/scripts/query_prometheus.py --query "up{job='api'}" --time-range 120

# Explicit datasource ID (use list_datasources.py to find it):
python .claude/skills/observability-grafana/scripts/query_prometheus.py --query "up" --datasource-id 8

Note: The Prometheus datasource ID is auto-detected by default. Use list_datasources.py if you need to target a specific datasource.

list_datasources.py - List Configured Datasources

python .claude/skills/observability-grafana/scripts/list_datasources.py

get_alerts.py - Get Active Alerts

python .claude/skills/observability-grafana/scripts/get_alerts.py

create_dashboard.py - Create Dashboard from Template

python .claude/skills/observability-grafana/scripts/create_dashboard.py \
  --title "My Dashboard" \
  --template .claude/skills/observability-grafana/templates/vercel-overview.json

# With folder:
python .claude/skills/observability-grafana/scripts/create_dashboard.py \
  --title "My Dashboard" \
  --template .claude/skills/observability-grafana/templates/vercel-overview.json \
  --folder-uid FOLDER_UID

# Overwrite existing:
python .claude/skills/observability-grafana/scripts/create_dashboard.py \
  --title "My Dashboard" \
  --template .claude/skills/observability-grafana/templates/vercel-overview.json \
  --overwrite

---

Dashboard Templates

Pre-built templates are in .claude/skills/observability-grafana/templates/:

Template	Description
vercel-overview.json	4-panel Vercel monitoring: error count, error rate, log stream, request summary

---

PromQL Quick Reference

# Request rate
rate(http_requests_total{service="api"}[5m])

# Error rate %
sum(rate(http_requests_total{status=~"5.."}[5m])) / sum(rate(http_requests_total[5m])) * 100

# P95 latency
histogram_quantile(0.95, rate(http_duration_seconds_bucket{service="api"}[5m]))

# CPU usage
rate(process_cpu_seconds_total{service="api"}[5m])

---

Investigation Workflow

1. list_datasources.py                    # Find available datasources
2. list_dashboards.py --query "api"       # Find relevant dashboards
3. get_dashboard.py --uid <uid>           # Get dashboard details
4. query_prometheus.py --query "rate(http_requests_total{service='api',status=~'5..'}[5m])"
5. get_alerts.py                          # Check active alerts