Excel / XLSX

**Judgement:** A strong, specialized Excel/XLSX skill focused on preserving real-world workbook behavior (formulas, dates, structure, and templates). Best suited for cases where the Excel file itself is the primary artifact and must remain a live, trustworthy model. **Key strengths** - Very clear guidance on when to use openpyxl vs pandas vs CSV, which reduces tool misuse. - Deep attention to Excel’s quirks: date serial systems, time as fractional days, 1900/1904 modes, and the 1900 leap-year bug. - Strong emphasis on keeping calculations as formulas in Excel, not hardcoded values, and on avoiding stale cached results. - Explicit focus on preserving formatting, merged cells, named ranges, validations, filters, conditional formatting, print areas, and sheet order—critical for template-based business workbooks. - Good coverage of common traps: type inference problems, ID/phone/ZIP corruption, long-number precision loss, off-by-one column errors, broken external references. **Main risks / limitations** - Relies on openpyxl, which does **not** recalculate formulas; users must ensure workbooks are recalculated in Excel or another engine before relying on values. - Large workbooks can still be slow or memory-heavy despite the recommended streaming and chunked-reading patterns. - Does not address every advanced Excel feature (e.g., deep macro/VBA manipulation, pivot table authoring), so some complex legacy models may be partially out of scope. - Cross-application differences (Excel vs Google Sheets vs LibreOffice) remain a risk; the skill highlights this but cannot fully eliminate it. **Recommended scenarios** - Preparing or updating deliverable Excel models (finance, FP&A, analytics, operations) where end users will work in Excel and require intact formulas and layout. - Filling and modifying corporate templates or regulatory forms where formatting, validations, print layout, and merged cells must be preserved. - Any workflow where data types must survive Excel’s auto-conversion (IDs, SKUs, phone numbers, ZIP codes, long IDs) without silent corruption. - Use together with `data` / `data-analysis` skills for heavy computation, reserving this skill for the Excel-specific modeling and presentation layer.