Excel / XLSX

• The task involves a .xlsx, .xlsm, .xls, .csv, or .tsv file and the output must be a well-formed Excel workbook.
• Formulas, cell references, named ranges, merged cells, or workbook structure must be preserved or created correctly.
• You are editing an existing template and need to match its visual conventions, sheet order, print settings, and validations.
• Data types are sensitive — long IDs, phone numbers, leading zeros, dates, or mixed columns that Excel could silently corrupt.

• You only need plain CSV export with no formatting or formula requirements — the csv skill is a lighter fit.
• The task is purely statistical analysis or data reshaping with no Excel delivery requirement — consider data or data-analysis instead.

The skill enforces writing live formulas into cells rather than hardcoding derived values, and flags stale cached results. It checks for #REF!, #DIV/0!, #VALUE!, #NAME?, and circular-reference errors before delivery, and warns that openpyxl preserves formulas without recalculating them.

Excel stores dates as serial numbers subject to the 1900 false-leap-day bug and the alternate 1904 date system. This skill handles both systems, ensures time values are treated as fractional days, and validates that number formats display the correct value to the end user.

Long identifiers, phone numbers, ZIP codes, and leading-zero values are stored as text to prevent silent truncation or reformatting. The skill explicitly handles mixed text-number columns and guards against scientific-notation conversion, auto-parsed dates, and precision loss past 15 digits.

Templates are treated as authoritative: sheet order, column widths, freeze panes, filters, print areas, data validation, conditional formatting, and merged-cell logic are all preserved unless the task explicitly changes them. Newly inserted cells match the surrounding style rather than introducing a foreign visual system.

The skill routes work to the right library: pandas for analysis and CSV-like reshaping, openpyxl for formula authoring, styling, merged cells, and workbook preservation. It treats CSV as a plain-data exchange format rather than a feature-complete Excel substitute.

Large workbooks are read with streaming or chunked strategies to avoid memory spikes and accidental full-sheet loads. Narrow reads, explicit dtypes, and sheet targeting keep large-file workflows from causing unintended damage.

An agent receives a multi-sheet financial model template and must update assumption cells and extend formula rows. The skill ensures formulas reference assumption cells (not hardcoded numbers), copied formulas are checked for relative-reference drift, and the workbook ships with zero formula errors.

A business user provides a branded .xlsx template with conditional formatting, print areas, and frozen headers. The skill populates data rows while preserving every visual convention — column widths, merged header cells, filters, and validation dropdowns — so the output opens correctly in Excel without cosmetic cleanup.

A dataset containing product SKUs, phone numbers, and ZIP codes must be written to Excel. The skill stores these as text to prevent Excel from stripping leading zeros or converting long IDs to scientific notation, and explicitly handles any mixed-type columns.

An agent audits an .xls or .xlsm file for stale cached values, broken external references, hidden sheets with active named ranges, and formula errors. The skill surfaces common traps — including macro-containing formats and 1900/1904 date-system mismatches — before any edits are saved.