Профилирование и анализ качества данных

★ 7.3 · content

explore-data is a Claude Code skill that generates a comprehensive data profile for a table or file — CSV, Excel, Parquet, or JSON — covering structure, quality, and patterns in one pass. It connects to a data warehouse via an MCP server or reads an uploaded file directly, then computes row and column counts, null rates, cardinality ratios, top and bottom value frequencies, and full numeric percentiles (p1 through p99). Every column is classified as an Identifier, Dimension, Metric, Temporal, Text, Boolean, or Structural field; columns with more than 5% nulls trigger a warning and more than 20% raise an alert. Duplicate detection, skewed distributions, inconsistent formatting, and placeholder values like "test" or "999999" are all flagged with severity levels. After profiling, the skill recommends 3–5 follow-up analyses — trends, cohorts, correlations, or quality investigations — making it the right starting point before choosing metrics for any dashboard or report.