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
- rittmananalytics/wire-plugintheir own version, first seen Oct 13, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- rittmananalytics/wire-plugin★ 14same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository21 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- analyzeSkill
- build-dashboardSkill
- build-modelSkill
- build-pipelineSkill
- causal-analysisSkill
- connect-dataSkill
- define-kpisSkill
- evalSkill
- experimentSkill
- forecastSkill
- health-checkSkill
- investigate-metricSkill
- make-deliverableSkill
- powerbiSkill
- privacy-reviewSkill
- research-domainSkill
- scorecardSkill
- setup-teamSkill
- triageSkill
- upgradeSkill
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
- Very specific: line 18
Spec before building: Do not re-derive requirements here - /build-dashboard steps 1-4 own the audience, the 1-3 questions, the metrics …
- Very specific: line 24
Ship a build script, not a workbook: Every dashboard folder gets a build.py that regenerates both the extract and the .twb from source.
- Distinctive: line 26
Validate before anyone opens it: Run the checker:
≈ “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.
Files
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
- 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. - Detect what you can actually do. Read
references/tooling-tiers.mdand 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. - Spec before building. Do not re-derive requirements here -
/build-dashboardsteps 1-4 own the audience, the 1-3 questions, the metrics (which must already exist inknowledge/metrics-catalog.md), the filters, and the layout sketch. This skill owns implementation only. If there's nodashboards/<name>/SPEC.mdyet, go write one first. - 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 perreferences/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 useextract=Trueand 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. - Write the spec, not the XML. You author a Python dict and
scripts/twb_builder.pyemits schema-correct XML. A worksheet is 150-250 lines of order-sensitive boilerplate; the builder owns that so you don't. The spec format isreferences/workbook-spec.md; the six chart recipes arereferences/chart-recipes.md. Chart choice, colour, layout, and decluttering come fromstandards/dashboard-standards.md- that file is the single source of truth for design and this skill does not restate or override it. - Ship a build script, not a workbook. Every dashboard folder gets a
build.pythat regenerates both the extract and the.twbfrom source. A workbook someone once saved is not reproducible; a script that rebuilds it is (CLAUDE.mdprinciple 3). - Validate before anyone opens it. Run the checker:
Fix every ERROR. Triage WARNs, don't ignore them. For anything stakeholder-facing, hand off topython .claude/skills/tableau/scripts/validate_twb.py dashboards/<name>tableau-validatorfor an independent pass. - 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 intodashboards/<name>/, update thedashboards/README.mdinventory. Publishing to Server/Cloud for a broad audience -> confirm with the user first. - 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 withencoding="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.twband 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/upgradecan 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/
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
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.

