Grafana Data Links and Drilldowns Prompt

You are a senior observability engineer who builds Grafana data links and drilldowns so a click on a series or cell carries context to the right dashboard, Explore query, or external runbook.

I will provide:
- The source panel and what a user wants to reach next
- The variables/labels available in the data
- The target (another dashboard, Explore, external URL)

Your job:

1. **Choose link type**: panel-level data link (applies to whole panel), field-override data link (per field/series), or a cell link on tables.
2. **Use built-in variables**: interpolate `${__value.raw}`, `${__field.name}`, `${__series.name}`, `${__data.fields[...]}`, and dashboard vars like `${namespace}` into the URL.
3. **Preserve time range**: append `${__url_time_range}` so the target opens on the same window; add `${__all_variables}` to carry every template variable.
4. **Link to another dashboard**: build `/d/<uid>/<slug>?var-namespace=${namespace}&${__url_time_range}`.
5. **Link to Explore**: construct a `/explore?left=` payload with the datasource and a query templated from the clicked value for ad-hoc drilldown.
6. **Field overrides**: use "Add field override" > "Data links" to attach links only to specific series (e.g. only the `pod` field opens the pod dashboard).
7. **External tools**: link to runbooks, logs, or trace UIs, URL-encoding interpolated values; open in a new tab where appropriate.
8. **Test interpolation**: verify variables resolve (no literal `${...}` in the final URL) across table, time series, and stat panels.

Mark DESTRUCTIVE: none normally — links are read-only navigation. Flag only links that trigger state-changing external actions (e.g. a link that hits a write API).

---

Source panel: [DESCRIBE]
Available variables/labels: [DESCRIBE]
Target: [DESCRIBE]

Why this prompt works

Grafana’s data-link variable system (${__value.raw}, ${__field.name}, ${__url_time_range}, ${__all_variables}) is powerful but under-documented, and the common failure is a link that drops the time range or leaves a literal ${...} in the URL. This prompt picks the right link scope, carries time and variable context, and forces interpolation testing so drilldowns actually land where the responder needs.

How to use it

1. Describe the click-to-destination flow you want.
2. List the fields/labels present so links reference real data.
3. Always carry the time range unless you deliberately reset it.
4. Test every panel type the link appears on.

Useful commands

# Fetch a dashboard to inspect existing data links
curl -s -H "Authorization: Bearer $TOKEN" \
  http://localhost:3000/api/dashboards/uid/svc-overview | jq '.dashboard.panels[].fieldConfig.defaults.links'

# Resolve a dashboard slug/uid for building links
curl -s -H "Authorization: Bearer $TOKEN" \
  "http://localhost:3000/api/search?query=pod%20detail" | jq '.[] | {uid,url}'

Example config

Panel field config with a data link into another dashboard and one into Explore:

{
  "type": "table",
  "title": "Pods by error rate",
  "fieldConfig": {
    "defaults": { "links": [] },
    "overrides": [
      {
        "matcher": { "id": "byName", "options": "pod" },
        "properties": [
          {
            "id": "links",
            "value": [
              {
                "title": "Open pod dashboard",
                "url": "/d/pod-detail/pod-detail?var-pod=${__value.raw}&var-namespace=${namespace}&${__url_time_range}",
                "targetBlank": false
              },
              {
                "title": "Logs in Explore",
                "url": "/explore?left=%7B%22datasource%22:%22loki-prod%22,%22queries%22:%5B%7B%22expr%22:%22%7Bpod%3D%5C%22${__value.raw}%5C%22%7D%22%7D%5D,%22range%22:%7B%22from%22:%22now-1h%22,%22to%22:%22now%22%7D%7D",
                "targetBlank": true
              }
            ]
          }
        ]
      }
    ]
  }
}

Common findings this catches

• Literal ${...} in URLs → interpolation not resolving on this panel type.
• Lost time range → ${__url_time_range} omitted.
• Blank target filters → variable absent on the destination dashboard.
• Wrong context → override attached to the wrong field.
• Broken links → hardcoded UID after target recreated.
• Leaked internals → ${__all_variables} sent to an external tool.

When to escalate

• Standardizing a triage navigation graph across dashboards — observability lead.
• Deep-links into external ticketing/write systems — owning team + security.
• Explore payload templating that must survive schema changes — platform team.