Original LinkedIn Post
From comment to code in 24 hours ? thank you Jeffrey Shaffer! Yesterday, Jeffrey Shaffer dropped a comment that completely changed how I think about AI-powered dashboard creation: "Have you experimented with specialized agent/'skills' workflow? Maybe something like 'calculation buildermd' then 'chart buildermd' then 'dashboard makermd' then 'formattingmd'..."
He was 100% right. So I built it. Today I'm shipping?cwtwb v0.7.0?with a brand new?Agent Skills System?? 4 expert-level skill files that transform how AI agents build Tableau dashboards: Phase 1: calculation_builder ? Parameters, LOD expressions, field naming Phase 2: chart_builder ? Chart type selection, encoding strategy Phase 3: dashboard_designer ? Layout design, filter panels, interactions Phase 4: formatting ? Colors, sorting, number formats, tooltips The key insight:?Without skills, AI knows?how to call the API. With skills, AI knows?how to think like a Tableau expert. Each skill is exposed as an MCP Resource (cwtwb://skills/{name}), so any AI agent can load domain expertise on demand ? no fine-tuning, no training data, just markdown. Even with a simple prompt like "build me a sales dashboard," the skills ensure: Bar charts are sorted (the #1 amateur mistake!) KPI bars get proper 40-60px fixed height Maps use continuous color encoding Filter sidebar follows the Date ? Region ? Detail hierarchy This is the power of the open-source community. One comment from a Tableau Visionary ? a new feature shipped to PyPI within a day. Thank you Jeffrey Shaffer for the inspiration! The Big Book of Dashboards is literally embedded in cdwb's DNA now. What other skills would you want to see??accessibilitymd??performancemd??storytellingmd? Let me know!
#Tableau #AI #MCP #DataVisualization #OpenSource #AgenticAI #DashboardDesign #cwtwb #tableau #bi #agent #datafam #dataviz #skills