Grafana Dashboards

Provides production-ready Grafana dashboards with consistent layout, safe queries, and operator-focused usability.

Use this skill when

• A request asks to create or improve Grafana dashboards
• A request asks to standardize dashboard layout for on-call usability
• A request asks for dashboard JSON templates or snippets

Do not use this skill when

• The request is for end-to-end observability architecture beyond dashboards
• The task is unrelated to Grafana dashboards

Required inputs

• Target service/domain and dashboard purpose
• Audience (on-call, developer deep dive, leadership KPI)
• Data sources available (Prometheus/Mimir, Loki, Tempo/Jaeger, etc.)
• SLOs or KPIs (if available)
• Existing dashboard JSON or screenshots (if refactoring)
• Constraints (time range defaults, label cardinality limits, naming standards)

Workflow (Deterministic)

1. Confirm scope and data sources.
   • Output: a 2-4 sentence scope summary + list of data sources.
   • Decision: if any required data source is unknown/unavailable, ask for it before continuing.
2. Select a layout template based on audience.
   • Output: row-by-row layout sketch with row intent.
   • Decision: if KPI-focused, add a KPI row before symptom signals.
3. Specify panels for each row.
   • Output: panel list with question, viz type, unit, threshold, and query stub.
   • Decision: if a panel depends on a missing metric, propose a fallback panel or mark it as "needs metric".
4. Draft queries and variables safely.
   • Output: query list + variable list with label constraints.
   • Decision: if a query risks high cardinality, recommend a recording rule or pre-aggregation.
5. Add drilldowns and links.
   • Output: link map to logs/traces/detail dashboards.
6. Produce dashboard JSON or snippets.
   • Output: Grafana JSON sections or template references.
7. Run quality gates and note fixes.
   • Output: pass/fail checklist with remediation steps.

Quality Gates

• The top row answers: "is it broken?"
• An on-call person can find a likely cause within 2-3 clicks.
• Queries are performant (recording rules for expensive aggregations).
• Panels are stable (avoid tiny denominators; avoid misleading averages).

Common pitfalls to avoid

• Using unbounded labels (wildcards or regex on high-cardinality labels).
• Relying on averages for latency or error rates without percentiles.
• Mixing multiple questions into a single panel.
• Omitting units or thresholds, which hides intent.
• Building dashboards that only work at one specific time range.

Assets (Copy/Adapt)

• Dashboard stubs:
  • assets/dashboard-templates.json
  • assets/api-dashboard.json
  • assets/infrastructure-dashboard.json
  • assets/database-dashboard.json
• Panel + templating snippets:
  • assets/panel-examples.json
• Alert rule patterns (structure only):
  • assets/alert-templates.json

Output contract

Return a report using this format and keep the section order:

1. Summary
2. Inputs & Assumptions
3. Layout Sketch (rows + intent)
4. Panel Specs (question, viz, unit, threshold, query stub)
5. Queries & Variables (safe label bounds)
6. Drilldowns & Links
7. JSON Snippets or Template References
8. Quality Gates (pass/fail + fixes)

Example (Input → Output)

Input: "Create an on-call Grafana dashboard for the payments API using Prometheus and Loki. Focus on latency, errors, and top routes."

Output (abridged):

1. Summary: On-call overview for payments API with symptom-first layout.
2. Inputs & Assumptions: Prometheus + Loki available; SLO not provided.
3. Layout Sketch: Row 1 symptoms; Row 2 top routes; Row 3 infra saturation + logs.
4. Panel Specs: Error rate (timeseries, %, threshold 1%); p95 latency (ms); RPS.
5. Queries & Variables: service="payments", route variable (top 20).
6. Drilldowns & Links: Loki logs filtered by service + route.
7. JSON Snippets: assets/dashboard-templates.json skeleton + panel JSON blocks.
8. Quality Gates: Pass; add recording rule for p99 latency if needed.

References (Optional)

• Index: references/README.md
• Design guide: references/dashboard-design.md
• Implementation playbook: references/implementation-playbook.md