Excel Spreadsheet Assistant

Problem

When generating or editing Excel files, the hard part is not opening the workbook; it is formulas without cached values, cells reading back as None, broken cross-sheet references, inconsistent finance-model formatting, and using functions the local runtime cannot evaluate. This skill structures XLSX work into creation, editing, bulk data access, and formula verification to avoid common openpyxl, pandas, and LibreOffice pitfalls.

How It Works

• Create/edit: Prefer openpyxl for cells, formulas, formatting, and .xlsm macros; use pandas read_excel and to_excel for bulk data.
• Read/model: markitdown gives a quick sheet view but no coordinates, so it should not drive edits; reading both formulas and cached values takes two load_workbook passes.
• Formula checks: Files with formulas must be recalculated. openpyxl writes formulas as strings, so before recalculation many tools see None; if recalculation reports errors_found, fix the reported cells and run it again.
• Finance models: Apply color conventions for inputs, formulas, cross-sheet links, cross-file links, and key assumptions, and use fixed formats for currency, percentages, years, and multiples. Assumptions should sit in labeled cells referenced by formulas.

Boundaries

Zero formula errors is not proof that the business result is correct. Recalculation only proves formulas evaluate. The skill avoids functions such as XLOOKUP, FILTER, and UNIQUE, preferring INDEX/MATCH and legacy functions. When editing an existing workbook, follow its conventions, leave existing formulas untouched, and do not add template rows to files that only need edits.