XLSX — Excel / spreadsheet creation

Create .xlsx Excel spreadsheets (or .ods / Google Sheet variant) from tabular content. Invoked by source-to-artifact when the user wants a spreadsheet — typical asks: report builder, data summary, financial model, list with formulas.

Inputs

From the caller:

• data source: workspace CSV / TSV / JSON / pasted table, OR structured data extracted from a document
• target format: xlsx (Excel native), ods (LibreOffice), gsheet (Google Sheet)
• filename: e.g. q3-sales.xlsx
• sheets: optional; if multiple sheets, structure per-sheet

Backend per format

xlsx (native, fast)

Use openpyxl in the workspace:

from openpyxl import Workbook
from openpyxl.styles import Font, PatternFill, Alignment
wb = Workbook()
ws = wb.active
ws.title = '<sheet name>'
ws.append([header1, header2, header3])
for h in ws[1]: h.font = Font(bold=True)  # bold header
for row in rows: ws.append(row)
# Formula example:
ws['D2'] = '=SUM(B2:C2)'
# Column widths:
ws.column_dimensions['A'].width = 25
wb.save('<filename>.xlsx')

For multiple sheets: wb.create_sheet('<name>').

Write to workspace. Attach to reply.

ods (via cerase-office-converter)

Create .xlsx natively, then:

call_recipe("cerase-office-converter.convert_xlsx_to_ods", {input_b64: <base64 of .xlsx>})

gsheet (via google-workspace MCP)

call_recipe("google-workspace.sheets_create", {
  title: "<filename without ext>",
  data: [[<row 1 cells>], [<row 2 cells>], ...],
})

Returns {sheet_id, sheet_url}. Surface URL.

Best practices

• Header row always styled: bold + background fill + frozen pane (openpyxl ws.freeze_panes = 'A2').
• Formulas, not hardcoded values: when summing / averaging / comparing, use =SUM(...) etc. The user can extend.
• Number formatting: dates as YYYY-MM-DD, currency with explicit symbol + 2 decimals (#,##0.00 "€"), percentages with 1 decimal (0.0%).
• Multiple sheets: use when the data has logical groupings (e.g. per-month, per-region). Avoid stuffing everything in one wide sheet.
• Chart: don't add unless explicitly asked — charts are fragile across export targets (ODS/Google handle them differently). If asked, use openpyxl BarChart / LineChart / PieChart.

Don't

• Don't write column widths blindly: estimate from content length (~1.2 char per width unit).
• Don't fabricate data: if a cell is unknown, leave it blank + add a note in a _notes column.
• Don't ignore the cerase-office-converter for cross-format — handles edge cases (frozen panes, conditional formatting) correctly during conversion.