Skip to main content
Spreadsheets are where a lot of “AI-generated content” demos quietly fall apart. The model produces a table that looks right, but the formulas are dead text and the chart is a screenshot. The xlsx skill removes both problems: every formula is live, the chart is a real chart object, and conditional formatting survives the round-trip into Excel or Google Sheets.

​
The prompt

The budget tracker prompt entered into chat.yambr.com

​
What happened

The xlsx skill wraps openpyxl plus a LibreOffice-based recalculation step. After generating the workbook, the model runs recalc.py on it to confirm every formula evaluates without #REF! or #DIV/0!. That validation step is the difference between “the model wrote a spreadsheet” and “the spreadsheet works.” The model’s own confirmation:
The recalculation script confirmed all 38 formulas work correctly with zero errors ✅

​
The output

Completed budget tracker — sheet 1 Budget visible in the right panel with real values What you can see in the right-hand preview:
  • Sheet 1 “Budget” with 8 categories and realistic values (Paid Search Ads 8,500/8,500 / 7,850 · Social Media Ads 12,000/12,000 / 11,200 · Events & Trade Shows 15,000/15,000 / 14,500 · …).
  • A TOTAL row at the bottom totalling 60,500.00/60,500.00 / 59,500.00.
  • A Summary sheet tab next to the Budget tab.
The reply itself spells out the structure:
  • 8 realistic marketing budget rows covering paid search, social media, content, events, software, freelancers, travel, tools.
  • Live formulas: Variance = Planned − Actual and % Used = Actual / Planned * 100.
  • Totals row with all formulas calculating across the board.
  • Currency formatting on dollar amounts.
  • Conditional formatting on Variance: green for over budget, red for under budget.
  • Sheet 2 — Summary Dashboard: four key metrics (Total Planned, Total Actual, Overall Variance, Average % Used) plus the Planned-vs-Actual bar chart.

​
Why it works

Three things have to be true for a spreadsheet output to be useful:
  1. Formulas have to be live — otherwise the user can’t change a value and watch the totals update.
  2. The chart has to be a real chart object — otherwise you can’t restyle it or extend the data range.
  3. Conditional formatting has to survive the format — otherwise the variance column is just plain text.
The xlsx skill handles all three, and the bundled recalc.py script catches formula errors before the file ever reaches the user. Next: Freelance invoice in PDF →