Tableau Governance Scanner
A Claude skill that audits Tableau Cloud or Server sites for governance and hygiene issues. Built for Claude Desktop and the Tableau MCP.
How It Works
Say "scan CCI Platform" or "audit my site" and the skill pulls your content via the Tableau REST API, runs checks across 7 domains, and returns a letter-graded health report with prioritized actions.
CCI Platform – Governance Scan
D- – 61.5/100
"Significant issues. Trust and discoverability degraded."
Staleness: D | Usage: B | Certification: D | Naming: D
Structure: D | Performance: A | Metadata: C
Quick Wins (raise to ~74):
1. Add tags to all 14 views – 5 min
2. Add descriptions to 2 workbooks – 10 min
3. Archive 3 zero-view workbooks – 5 min
Long-Term Goals:
1. Certify primary datasource (8 observed downstream workbooks, sampled)
2. Fix stale pipeline (users on 99-day-old data)
3. Add co-owner (bus factor = 1)
Audit Domains
| Domain | What It Catches |
|---|---|
| Staleness | Content not updated within threshold |
| Usage & Adoption | Ghost workbooks, zero-view content, dormant legacy dashboards |
| Certification & Trust | Uncertified datasources with high downstream impact |
| Naming | "Copy of", "Sheet 1", "(old)", test content in production |
| Structure | Default project clutter, owner concentration, sandbox sprawl |
| Performance | Over-sheeted workbooks, large files, live connections under load |
| Metadata | Missing descriptions, low tag adoption |
Output Formats
| Format | Use Case |
|---|---|
| Control Tower | Interactive HTML dashboard with filters and severity gauges |
| Spreadsheet | Shareable .xlsx with one row per finding |
| Word | Formal .docx report for leadership |
| Markdown | Lightweight, good for wikis or chat |
| JSON | Programmatic consumption |
Presets
| Preset | For |
|---|---|
| Light | First scan or small site. Flags only clear problems. |
| Standard | Balanced. The default. |
| Strict | Mature orgs with governance teams. |
Key Features
- Zero-config execution. Infers scope, preset, and format from natural language.
- Compare mode. Side-by-side project comparison or delta tracking over time.
- Memory-aware. Remembers config per scope and preset, and shows score changes between comparable scans.
- Graceful degradation. Proceeds when datasource/project APIs return 401/403, grades skipped domains N/A, and labels the run Partial.
- Validation phase. Arithmetic, deduplication, and false-positive checks before output.
- Scheduled monitoring. Set it and forget it. Get alerted only when scores drop or new issues appear.
Note on permissions: Scheduled runs may prompt for tool approval the first time they access Tableau MCP endpoints. Run the task manually once ("Run now") to pre-approve the tools it needs. After that, future runs execute without interruption.
Requirements
- Claude Desktop with the Tableau MCP connected
- Tableau Cloud or Server REST API access
- Minimum permissions:
list-views+list-workbooks(datasource/project access optional)
Installation
- Save the whole
tableau-governance-scannerfolder (SKILL.mdplus theskill-expertise/files) to your Claude skills directory. InstallingSKILL.mdalone leaves out the check catalog and output formats it reads. - Trigger with: "audit my site", "governance scan", "scan [project name]", or "what needs attention?"
Companion Skills
| Skill | Handoff |
|---|---|
| VizCritique Pro | Evaluate visual design of flagged views |
| Tableau Scribe | Auto-generate docs for undocumented content |
| Data Quality Sentinel | Profile data in flagged datasources |
| Dashboard Architect | Plan redesigns for monolith workbooks |
Limitations
- No permission auditing (not exposed by the Tableau MCP)
- No extract refresh schedules (uses
updatedAtas proxy) - Downstream workbook counts for datasources are sampled (up to 10 workbooks) and are a lower bound
- No render-time profiling (sheet count and size are proxy signals)
totalViewCountis lifetime cumulative. Skill uses recency-adjusted logic to compensate.