Проверка аналитики перед презентацией стейкхолдерам

★ 7.3 · content

validate-data is a Claude Code skill that QA-reviews an analysis before it reaches stakeholders — catching calculation errors, methodology flaws, and visualization issues, including classic pitfalls like join explosion, survivorship bias, and shifting denominators. The eight-step workflow covers question framing, data source verification, aggregation and SQL spot-checks, chart review for misleading axes or scales, and an assessment of whether conclusions are actually supported by the evidence. It accepts SQL queries and results, markdown reports, notebooks, spreadsheets, or a plain description of methodology. The output is a structured Validation Report with a three-level readiness rating (Ready to share / Share with caveats / Needs revision), prioritized issues, required stakeholder caveats, and actionable improvement suggestions — making it the go-to skill for analysts and data engineers before any high-stakes presentation or report delivery.