excel

# excel Use this skill when the user asks for an Excel workbook or an `.xlsx` artifact, including creating, modifying, formatting, analyzing, validating, or exporting workbook content. For read-only analysis or validation requests, inspect the workbook and report findings without writing a new file unless the user asks for one. In tenant runtimes, new workbook generation is productized through `skills/excel/runtime/renderer.py`. Generate the workbook locally in the tenant workspace, verify it, then publish it with the `emit_artifact` tool using: - `kind: "spreadsheet"` - `content_encoding: "base64"` - `filename: "<safe-name>.xlsx"` - `mime_type: "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"` - `destinations: ["ui", "chat"]` ## Core rules - Produce a real `.xlsx` workbook. Do not rename CSV, TSV, HTML, JSON, or plain text files to `.xlsx`. - Use `runtime/sanitize.py` at the cell-write boundary. Strings starting with `=`, `+`, `-`, or `@` are text unless the cell is explicitly typed as a formula. - Preserve source data. Keep raw inputs available when appropriate, such as a `Source Data` or `Raw Data` sheet, and avoid silently replacing user-provided values with summaries. - Use workbook features that fit the task: multiple sheets, formulas, number formats, styles, column widths, freeze panes, filters, charts, data validation, named ranges, and notes/comments when they materially improve usability. - Keep formulas as formulas unless the...