XLSX Excel Spreadsheet Processing
Claude reads, cleans, and generates Excel spreadsheets — and is required to use live formulas instead of hardcoding computed numbers
What it does & when to use it
Anthropic's official Excel-handling skill. Beyond basic read/write, its most notable feature is a full financial-modeling style guide — blue/black/green/red color-coding conventions, number formatting rules, and one hard requirement: every calculation must be expressed as an Excel formula, never computed in Python and hardcoded into a cell, or the sheet silently breaks the moment an input changes.
Reads, cleans, and analyzes spreadsheet data with pandas; generates new spreadsheets with formulas, styling, and column widths via openpyxl; enforces that all calculations use Excel formulas rather than hardcoded values, so the sheet stays recalculable when inputs change; recalculates formulas and detects errors (#REF!, #DIV/0!, #VALUE!, #NAME?) with the bundled recalc.py script (requires LibreOffice); and for financial-model contexts, defines a color-coding convention (blue = hardcoded input, black = formula, green = same-workbook reference, red = external-file reference, yellow fill = assumption needing attention) plus number-format rules (years unformatted, currency as $#,##0, negatives in parentheses, etc.).
- Clean up a messy raw CSV into a properly structured Excel spreadsheet
- Build a financial model (e.g. a revenue forecast) where every cell is a formula and changing one assumption flows through the whole sheet
- Audit an existing Excel file for #REF!/#DIV/0! formula errors and fix them
- Fill in a new period's data into an existing template while preserving its formatting conventions
- Analyze tabular data and produce a summary sheet with charts
Pros & cons
- The "formulas, never hardcoded values" rule is the standout feature — it's the exact shortcut most code-generated spreadsheets take (compute in Python, write the number in), producing a sheet that looks right but silently breaks the moment any input changes; this skill makes that a hard requirement
- Ships an actual financial-modeling color/format convention (blue inputs, black formulas, parenthesized negatives) that's genuinely useful for finance/banking-adjacent use cases
- recalc.py actually recalculates formulas via LibreOffice and checks for errors — it doesn't just generate a file and assume it's correct
- Licensed as Proprietary (source-available) — no free redistribution or commercial repackaging
- Formula recalculation depends on a local LibreOffice install, which needs extra setup on a typical desktop environment
- The financial color/format conventions are industry-specific and may not fit non-finance use cases — the skill itself says to defer to an existing template's conventions when one is present
How to install
Three options:
- Claude Code:
/plugin marketplace add anthropics/skills, then/plugin install document-skills@anthropic-agent-skills(bundled with pdf, pptx, docx) - Claude.ai: enabled by default on paid plans
- Claude API: upload via the Skills API per Anthropic's docs
How to use
Just describe the task — mentioning Excel, a spreadsheet, .xlsx, or asking to clean/analyze/generate tabular data triggers it automatically. E.g. "Turn this sales data into an Excel sheet with monthly summary formulas," or "Check this financial model for formula errors."
Compared to similar skills
Compared to a plain pandas/openpyxl script, the core value is two easily-skipped hard rules: insisting on live formulas instead of hardcoded numbers, and mandatory recalculation-and-verification after generation. Both get skipped in "good enough to run once" scripts, but they're essential for a spreadsheet meant to be reused and recalculated as inputs change. Among the sibling pdf/pptx/docx skills, xlsx is the only one that's really about computational correctness rather than layout.