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

tableau

Skill4 starsChanged 6 days agoMITView on GitHub

Build and edit Tableau dashboards as code via .twb workbooks - schema-correct XML generated from a compact spec, validated against Tableau's official XSD before anyone opens it. Use when the team's BI tool is Tableau and a dashboard, sheet, or data connection needs building, fixing, or reviewing.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Spec before building: Do not re-derive requirements here - /build-dashboard steps 1-4 own the audience, the 1-3 questions, the metrics …

    line 18
  2. Very specific:

    Ship a build script, not a workbook: Every dashboard folder gets a build.py that regenerates both the extract and the .twb from source.

    line 24
  3. Distinctive:

    Validate before anyone opens it: Run the checker:

    line 26

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Tableau - .twb Authoring & Validation

Owner: dashboard-developer, with tableau-validator for the validation gate and analytics-engineer for the data layer. Args describe the work, e.g. /tableau build the exec revenue dashboard.

A .twb is plain XML. That makes a Tableau dashboard an ordinary code artifact - spec'd, built, diffed, reviewed, and version-controlled like everything else the team ships. Since February 2026 Tableau publishes official XSD schemas for the format, explicitly as a reference for agents to build and validate against. Work it as code, not as a GUI you can't reach.

Procedure

  1. Check this is the right path. This skill applies when {{BI_TOOL}} is Tableau, or the user explicitly asked for Tableau. If the team's tool is Power BI, Looker, Excel, or nothing yet, stop and return to /build-dashboard - none of the guidance below transfers. If {{BI_TOOL}} is still an unfilled placeholder, ask which tool the team uses before building anything.
  2. Detect what you can actually do. Read references/tooling-tiers.md and run the detection commands. State the tier before you start. Tier 1 needs only Python and the free Tableau Public app; it is the honest default and it is enough to build a complete, correct dashboard. Never install anything yourself - recommend, state the licence cost, let the user decide.
  3. Spec before building. Do not re-derive requirements here - /build-dashboard steps 1-4 own the audience, the 1-3 questions, the metrics (which must already exist in knowledge/metrics-catalog.md), the filters, and the layout sketch. This skill owns implementation only. If there's no dashboards/<name>/SPEC.md yet, go write one first.
  4. Build the data layer first. Every sheet reads from a mart or summary table - no heavy logic in the workbook (standards/tableau-standards.md). Decide the connection per references/connections.md: a CSV exported from the mart (portable, driver-free) or a live warehouse connection with custom SQL. Targeting Tableau Public means an extract - it is extract-only and cannot open a live connection, so use extract=True and ship a .twbx. Tableau Desktop needs neither. One grain per workbook. Most broken Tableau dashboards are a grain problem wearing a chart's clothing - if revenue is monthly and tickets are per-ticket, you cannot put both in one flat extract without duplicating one of them.
  5. Write the spec, not the XML. You author a Python dict and scripts/twb_builder.py emits schema-correct XML. A worksheet is 150-250 lines of order-sensitive boilerplate; the builder owns that so you don't. The spec format is references/workbook-spec.md; the six chart recipes are references/chart-recipes.md. Chart choice, colour, layout, and decluttering come from standards/dashboard-standards.md - that file is the single source of truth for design and this skill does not restate or override it.
  6. Ship a build script, not a workbook. Every dashboard folder gets a build.py that regenerates both the extract and the .twb from source. A workbook someone once saved is not reproducible; a script that rebuilds it is (CLAUDE.md principle 3).
  7. Validate before anyone opens it. Run the checker:
    python .claude/skills/tableau/scripts/validate_twb.py dashboards/<name>
    Fix every ERROR. Triage WARNs, don't ignore them. For anything stakeholder-facing, hand off to tableau-validator for an independent pass.
  8. Reconcile, then ship. Open the workbook in Tableau Desktop or Tableau Public and confirm it renders - validation proves the files are well-formed, only Tableau proves the dashboard works. Cross-check every displayed number against an independent direct query, saved in dashboards/<name>/checks/. Screenshot into dashboards/<name>/, update the dashboards/README.md inventory. Publishing to Server/Cloud for a broad audience -> confirm with the user first.
  9. Record what you learned. New source quirks -> knowledge/data-sources.md. Methodological choices (exclusions, LOD semantics, fiscal handling) -> knowledge/decision-log.md. Anything materially off-trend you noticed while reconciling -> the Observations section there too.

Hard rules

  • XSD-valid does not mean it opens. Tableau says so directly: the schemas do not cover "connection attributes, calculated field contents, and references to workbook elements". The validator's semantic layer covers much of that gap, but a human opening the file is still the last gate. Never report a dashboard as working because validation passed.
  • Write every file as UTF-8 without BOM, with LF endings. On Windows, PowerShell's Set-Content, Out-File, and > all add a BOM. Write from Python with encoding="utf-8", newline="\n" - write_twb() already does.
  • Never hand-edit a .twbx. It is a zip archive, and the official schema explicitly does not cover packaged workbooks. Work the .twb and ship the data alongside it.
  • Never invent a metric definition in a calculated field. Calculated fields are for presentation logic. A metric defined in a workbook is invisible to the catalog and to every other workbook, which is exactly how two dashboards start disagreeing. Route conflicts to metrics-steward.
  • Never edit the vendored XSD in schemas/. It stays byte-identical to upstream so /upgrade can diff it against a new Tableau release. The two shim files beside it are ours and are documented in place.
  • Close Tableau before regenerating a workbook it has open. It does not watch the filesystem and will overwrite your rebuild from its in-memory copy on the next save.

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.