Stop getting markdown tables back — use the xlsx skill to generate real spreadsheets

The xlsx skill is a document-generation capability bundled into Cowork that writes Microsoft Excel files instead of returning tabular data as a markdown code block. When you say the word "spreadsheet", "Excel", or ".xlsx", it fires automatically — no slash command, no explicit invocation.

The difference matters once the output is a deliverable. A markdown table is fine for exploration, but if you're handing something to a finance team, importing into a reporting system, or building an incident tracker someone will actually maintain, you need a real file. The skill bridges that gap: describe the structure you want, and it generates an .xlsx with openpyxl under the hood — column headers, data types, number formatting, formulas, and optionally charts, all in one pass.

Incident cost tracking. SRE teams often track incidents in ad-hoc spreadsheets that get rebuilt every quarter. Ask the skill for an incident log with columns for date, service, severity (P1/P2/P3), downtime in minutes, estimated cost per minute, and a computed total cost column. Tell it to add a summary row at the bottom that counts incidents by severity and sums total cost. The skill handles the formula wiring — =SUMIF, =COUNTIF, conditional formatting that turns P1 rows red — without you specifying cell addresses.

Capacity planning grids. Upload a JSON or CSV export from your monitoring system — nodes, CPU utilization, memory used, disk free — and ask for an xlsx back with conditional formatting that highlights any cell above 80% utilization in amber and above 95% in red. The skill reads the input data, builds the sheet, and applies the rules. You get a file ready to share, not a description of how to build one in Excel manually.

Data cleaning from malformed sources. The skill also handles the reverse direction: you give it a messy CSV (misaligned headers, mixed date formats, extra blank rows) and describe the clean shape you want, and it outputs a properly typed xlsx. This is useful when you're pulling exports from legacy systems that can't be trusted to produce clean tabular data. The skill reads the source, normalises it, and writes the target format in a single step.

Where not to reach for it. Don't invoke xlsx for quick tabular answers in chat. If you're asking "what are the top five services by error rate this week?" and the answer fits in a few rows, a markdown table is faster and doesn't leave a file to manage. The xlsx skill earns its keep when the output is something you'll open in Excel or Numbers, share with a stakeholder, or import into another system — the deliverable case, not the exploration case. Using it for everything produces file sprawl and slower answers.

The skill also covers reading existing spreadsheets: attach an xlsx to your Cowork session and ask questions about it, or tell the skill to add a column, fix a formula, or reformat a range. The read path is as useful as the write path when you're iterating on a file that already exists.

The try-it block below shows a one-prompt demo that fires the skill, generates a complete incident log template, and saves it to your connected folder.