dsh-plugin-csv-report

In tariff data analysis, the data preparation stage that comes before any operational decision is often the most painful part: raw files arrive in inconsistent formats, analysis scripts are hard to reuse, quality issues are difficult to trace, and results lack a hash chain linking upstream and downstream artifacts. dsh-plugin-csv-report packages this entire workflow into two callable DeepSeek Harness tools, turning the path from raw tables to an auditable statistical report into a standard, reproducible, and traceable pipeline.

On the capability side, tariff_prepare reads XLSX, XLSM, and CSV files, standardises user identifiers, entry dates, and numeric fields, generates a 19-field data dictionary, and runs comprehensive quality checks covering required fields, duplicate identifiers, invalid values, brand enumerations, binary flags, and saturation ranges before emitting a normalised CSV and a preparation manifest. tariff_describe then produces a per-field data-quality report, computes mean, standard deviation, median, and quartiles for numeric fields, delivers Top-N distributions for categorical fields, and aggregates group statistics by plan name, user region, and brand series, culminating in a full Markdown report index.

The plugin is well suited for telecom operators or data teams working with de-identified tariff data in local environments. It does not make business decisions or perform causal inference; instead, it ensures that every step is connected by a SHA-256 hash chain spanning raw files, normalised data, and the final report, making results auditable and independently verifiable.