spreadsheets
Originally fromhmbown/deepseek-tui
Installation
SKILL.md
Spreadsheets
Opinionated tabular-data handling for macOS. TSV/CSV is the primary format; .xlsx is the exception. Stack: qsv for fast validation/profiling, duckdb for SQL, uv-run Python (stdlib csv + Decimal) for precision transforms, qsv excel/DuckDB/fastexcel for values-only Excel reads, XlsxWriter for new workbooks, openpyxl for existing workbook mechanics, and headless LibreOffice for formula recalculation. Run all Python through uv — never bare python or pip.
Tool Selection
scripts/peek.py, scripts/profile.py, and scripts/recalc.py are bundled with this skill. Resolve them relative to this SKILL.md's directory, not the user's current project. They are not supposed to exist in the target repo.
| Job | Use |
|---|---|
| First look or structural validation of CSV/TSV | uv run scripts/peek.py <file> [--strict] |
| Quality profile of CSV/TSV | uv run scripts/profile.py <file> [--markdown] |
| Counts, stats, frequencies, dedupe, column select | qsv (use --cache-threshold 0 for stats) |
| Joins, group-bys, pivots, cross-file SQL, format conversion | duckdb -c "..." |
| Row-level transforms, precision-critical edits | uv run Python with a PEP 723 header |
Anything .xlsx in or out |
read references/xlsx.md first |
Recalculating .xlsx formulas |
uv run scripts/recalc.py <file.xlsx> |
| Row/column-aware diff of two tables | bunx daff old.tsv new.tsv |
| Interactive viewing (suggest to the user; never launch TUIs) | csvlens, vd, Numbers.app |