agentleFS
link7373link7373/agentic-bi-team/.claude/skills/build-dashboard/SKILL.md

build-dashboard

Skill4 starsChanged 7 days agoMITView on GitHub

Spec and build a dashboard in the team's BI tool (Tableau, Power BI, Looker, or file-based) — requirements, data layer, layout, visual design, validation, and publishing. Use for new dashboards or redesigns.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Build Dashboard

Owner: dashboard-developer, with analytics-engineer for the data layer. Args describe the dashboard, e.g. /build-dashboard executive revenue overview.

Procedure

  1. Challenge the request (briefly). Is this a recurring monitoring need (dashboard) or a one-off question (/analyze is cheaper)? Does an existing dashboard in dashboards/README.md already cover 80% of it? Proceed once it's genuinely a dashboard need.
  2. Requirements. Identify: audience (from knowledge/stakeholders.md), the 1–3 questions it must answer, metrics (must exist in knowledge/metrics-catalog.md — route new ones through /define-kpis), filters/drill paths, refresh cadence, access level.
  3. Spec. Write dashboards/<name>/SPEC.md: all of the above plus a text/ASCII layout sketch following standards/dashboard-standards.md (headline number top-left, supporting trends, detail below the fold). Pick each chart from the selection/charts-to-avoid guidance there (length/position over area; no pies for comparison; colour-blind-safe; bars from zero — fuller rationale in analytics.md Part 2). For stakeholder-facing dashboards, confirm the spec with the user before building.
  4. Data layer. Every tile reads from a mart or summary table — no heavy logic in the BI tool. Missing tables → build via /build-pipeline first. Confirm the source tables refresh at least as often as the dashboard promises. Check each metric's classification in the catalog against the dashboard's audience, and every filter and drill path against the minimum group size in knowledge/governance-policy.md — a dashboard is where a harmless aggregate becomes a list of three people after two clicks. No restricted column reaches the canvas, a tooltip, or an export; a wider audience for personal data than before goes through /privacy-review.
  5. Build in {{BI_TOOL}}.
    • Tableau: hand off to /tableau, which builds the real artifact as a .twb workbook from a compact spec - schema-correct XML, .hyper extract where the target needs one - and validates it against the schema Tableau actually applies before anyone opens it. Extracts for slow sources; calculated fields for presentation only, never business definitions; details in standards/tableau-standards.md.
    • Power BI: hand off to /powerbi, which builds the real artifact as a PBIP project — TMDL semantic model, PBIR pages and visuals, theme JSON — and validates it before it reaches Desktop. Star schema, thin DAX named per catalog, dedicated date table; details in standards/powerbi-standards.md.
    • Looker: LookML in dashboards/lookml/, measures defined once at model layer, business-language labels.
    • No BI tool configured: deliver a self-contained HTML dashboard (vanilla JS + a charting lib, or static matplotlib images) regenerated by a script in dashboards/<name>/.
    • If direct tool access (API/MCP/CLI) isn't available, produce the complete artifact that can be imported (e.g. LookML files) plus exact setup instructions. This no longer applies to Power BI or Tableau — PBIP and .twb are both plain text, so build the project itself rather than describing it.
  6. Validate. Reconcile every displayed number against an independent direct query (save reconciliation queries in dashboards/<name>/checks/). Test filters, empty states, and the slowest interaction. Five-second test on the layout.
  7. Ship. Screenshot to dashboards/<name>/, update dashboards/README.md inventory (name, audience, source tables, refresh, owner, link). Publishing to a broad or external audience → confirm with the user first.
  8. Maintain. Note in the spec when the dashboard should be reviewed for retirement ({{DASHBOARD_REVIEW_CADENCE e.g. quarterly}}). Dead dashboards erode trust in all dashboards.

Created by Colin Beck — https://www.linkedin.com/in/beckcolin/

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.