Dashboard Forge
Generate Grafana dashboards from "I want to monitor my API" instead of writing PromQL by hand
Describe what you want to monitor and get a complete Grafana dashboard with proper PromQL queries, sensible thresholds, and standard layout. Also audits existing dashboards for accuracy and unused panels.
INGREDIENTS
PROMPT
Create a skill called "Dashboard Forge". Generate and manage Grafana dashboards: Mode 1 — Generate: When I describe what I want to monitor, create a complete Grafana dashboard JSON with: - Proper PromQL queries with correct functions (rate, histogram_quantile, etc.) - Standard panels: time series, stat, gauge, table as appropriate - Template variables for namespace, environment, pod selection - Sensible refresh rate and time range defaults - RED metrics (Rate, Errors, Duration) for services - USE metrics (Utilization, Saturation, Errors) for infrastructure Mode 2 — Audit: Given existing dashboard JSON or a Grafana URL I can access, review for: - Broken queries (referencing metrics that don't exist or labels that don't match) - Missing metrics (standard metrics not being graphed) - Layout and usability improvements - Alert rule suggestions for key panels Output as importable JSON and optionally as Grafonnet/Terraform code.
How It Works
Creating a good Grafana dashboard requires PromQL fluency, knowledge of
what metrics your services expose, and an eye for layout. This skill
handles all three — you just describe what you want to see.
What You Get
- Complete Grafana dashboard JSON from natural language descriptions
- Standard RED (Rate, Errors, Duration) and USE (Utilization, Saturation, Errors) dashboards
- Proper PromQL queries with correct label filters, rate windows, and aggregations
- Variables for environment, namespace, and service selection
- Alert rules embedded in dashboard panels
- Dashboard-as-code output (Grafonnet or Terraform)
Setup Steps
- Describe what you want to monitor (e.g., "my API's latency, error rate, and throughput")
- Tell your Claw what metrics are available (or let it query Prometheus to discover them)
- Review the dashboard JSON and import into Grafana
- Customize thresholds and layouts as needed
Tips
- Start with RED metrics for request-driven services and USE for infrastructure
- Include template variables so one dashboard covers all environments
- Keep dashboards focused — one per service, not one giant dashboard
- Ask for Grafonnet output if you want dashboards-as-code in version control
- Request an audit of existing dashboards to find stale or broken panels