Skip to main content

Tableau Governance Scanner for Codex

An evidence-first Codex skill for governance audits and explicitly requested remediation of Tableau Cloud and Tableau Server through a connected Tableau MCP server.

What it does​

  • Inventories visible projects, workbooks, views, and published data sources.
  • Reviews modification age, adoption evidence, trust and lineage, naming, organization, ownership concentration, performance-risk indicators, and metadata completeness.
  • Produces compact findings with stable IDs, observed evidence, calibrated inference, and next steps.
  • Supports project/site comparisons and user-approved baselines.
  • Uses normalized scoring so larger sites are not penalized merely for containing more content.

What it does not do​

  • It does not equate content modification time with data freshness.
  • It does not treat cumulative view counts as recent adoption.
  • It does not claim measured performance from structural proxies.
  • It does not audit row-level data quality, visual design, or permissions completeness.
  • It does not mutate Tableau content during a scan.

Requirements​

  • Codex with a connected Tableau MCP server.
  • Read access to the intended Tableau scope.
  • Optional artifact capabilities for XLSX, DOCX, or HTML outputs.

The available Tableau MCP implementation determines which domains can be assessed. The skill discovers tool schemas at runtime and reports coverage gaps.

Install​

Place the tableau-governance-scanner directory in a Codex skills location, then restart or refresh Codex so it discovers the skill.

Example prompts​

Use $tableau-governance-scanner to audit the Finance project with the Standard policy.
Run a read-only Strict governance scan for the visible Tableau site and return JSON plus a concise executive summary.
Compare these two Tableau projects for governance health. Do not rank them if coverage differs.

Package contents​

  • SKILL.md — Codex operating instructions.
  • agents/openai.yaml — UI metadata and Tableau MCP dependency.
  • skill-expertise/methodology.md — policies, domains, severity, and normalized scoring.
  • skill-expertise/mcp-and-evidence.md — capability discovery and evidence semantics.
  • skill-expertise/reporting.md — report, JSON, comparison, and baseline contract.
  • skill-expertise/score_governance.py — deterministic validation and scoring helper.
  • skill-expertise/test_score_governance.py — unit tests.

Scoring helper​

python skill-expertise/score_governance.py scan.json --pretty

The helper validates scoring inputs, excludes partial/failed entities, rejects duplicate findings, and calculates item-, domain-, and overall scores. It is arithmetic-only: the caller must independently validate evidence and assessed coverage.

Safety​

Audits remain read-only. Explicitly requested, bounded remediation can follow in the same run when the runtime exposes the necessary tools. Existing authorization is sufficient; ask only for unresolved scope or impact. Destructive actions are never inferred from audit findings.

Validation status​

Local structure, unit, and static checks can validate the package. A true forward test requires a separate Codex session connected to Tableau MCP; do not interpret local tests as proof of live connector compatibility.

License​

No license is asserted by this conversion. Add one only if you have the right to do so.