Create, repair, and recalculate spreadsheet workbooks without breaking formulas
Use the Anthropic xlsx skill when an agent needs to create, clean up, or modify .xlsx, .xlsm, .csv, or .tsv files as spreadsheet deliverables, not just inspect tabular data. It pushes the agent toward formula-safe edits, workbook validation, and recalculation instead of hardcoded outputs or one-off scripts.
Use the Anthropic xlsx skill when an agent needs to create, clean up, or modify .xlsx, .xlsm, .csv, or .tsv files as spreadsheet deliverables, not just inspect tabular data. It pushes the agent toward formula-safe edits, workbook validation, and recalculation instead of hardcoded outputs or one-off scripts.
npx skills add agentskillexchange/skills --skill create-repair-and-recalculate-spreadsheet-workbooks-without-breaking-formulas
The upstream tool here is Anthropic’s xlsx skill from the anthropics/skills repository. This entry is not a generic spreadsheet product card. It describes a bounded agent workflow: create, repair, restructure, and recalculate spreadsheet workbooks while preserving formulas, formatting intent, and deliverable quality. The agent behavior is concrete. It should pick spreadsheet-native operations, use formulas instead of hardcoded computed values, recalculate the workbook after edits, and verify that the file ships without Excel errors like #REF!, #DIV/0!, or broken cross-sheet references.
Invoke this skill when the user wants the output to be a spreadsheet file or wants an existing workbook fixed. Good fits include adding columns, cleaning malformed rows, rebuilding formulas, standardizing assumptions, exporting structured data into Excel, or rescuing a model that was damaged by unsafe edits. This is the right path when a normal product workflow would leave the user manually rebuilding formulas or checking every tab by hand.
The scope boundary matters. This is not a listing for Excel, pandas, or a generic data library. It is also not the right skill when the real deliverable is a dashboard, database pipeline, HTML report, or Google Sheets integration. The job-to-be-done is specifically workbook-safe agent execution. Integration points include Python data tooling such as pandas and openpyxl, LibreOffice-based recalculation, CSV/TSV ingestion, and downstream export to finished spreadsheet artifacts.