xlsx

name	xlsx
description	Create, edit, analyze, or clean Excel/.xlsx (and .csv) spreadsheets — formulas, multiple sheets, charts, formatting, data cleaning. Use when the user wants to read, build, or transform a spreadsheet.
license	original (see skills/ATTRIBUTIONS.md)
source	original clean-room; uses openpyxl (MIT); optional pandas (BSD-3), user-installed

Spreadsheets (.xlsx / .csv)

Pick the tool by task: openpyxl for cell-level control, formulas, styles, charts; pandas for bulk analysis/cleaning of tabular data.

Requirements (preflight)

python3 -c "import openpyxl" 2>/dev/null || python3 -m pip install openpyxl
# only if doing bulk analysis/cleaning:
python3 -c "import pandas"  2>/dev/null || python3 -m pip install pandas

Not bundled — install on demand, respect an active venv, and if install fails say so rather than faking results.

Inspect before you touch

Always characterize the workbook first: sheet names, dimensions, header row, column types, NaN/blank counts. Print this. Editing a spreadsheet blind is how data gets silently corrupted.

Editing with openpyxl

• load_workbook(path) (add data_only=True to read computed values instead of formula strings — know which you need).
• Address cells explicitly; preserve existing formatting/sheets you aren't changing.
• Write formulas as strings (ws["C2"] = "=A2*B2"); openpyxl does not compute them — Excel/LibreOffice does on open. If the user needs the computed value now, compute it in Python and write the value (and say which you did).
• Charts via openpyxl.chart; number/date formats via cell number_format.

Analysis / cleaning with pandas

read_excel/read_csv → operate → write back. For cleaning: dedupe, fix dtypes (esp. dates/IDs read as floats), normalize text/encoding, handle missing values deliberately (don't silently drop rows — report counts). Round-trip column order and types intentionally.

Discipline

• Output to a new file unless told otherwise; never overwrite source data without confirmation — spreadsheet edits are easy to get subtly wrong and hard to undo.
• After writing, reopen and verify row counts / a few known cells; report what changed (rows in/out, columns added). $ARGUMENTS is the file and/or task.