A CSV is a text file where each line is a row and the values of each column are separated by commas. Any spreadsheet, database or program can open it, so it works for moving data from one system to another without tying you to any of them. It is simple, but it has traps with separators, commas and accents.
A question from the app. Pick an answer.
The skill must always deliver the same CSV: same columns, same order. What do you do?
The description helps select when to use the skill, not how the CSV comes out.
Pep talks do not sort columns. 😅 Work that must come out the same every time needs another approach.
Repeatable, exact work goes in a script: it fixes the columns and their order. Rows depend on the orders.
Opened in a text editor, a sales CSV looks like this: a first line with the column names and, below it, one sale per line. No colors, formulas or tabs: just the data. That plainness is exactly its strength.
When you switch billing systems, send your sales to an accountant, or upload your customers to a new tool, the bridge is almost always a CSV. It is also the most direct way to give data to data science or to an AI.
fecha,cliente,producto,total
2026-11-02,Ana,Plan mensual,29.90
2026-11-03,Luis,Plan anual,299.001) The separator. In many countries the comma is the decimal mark (29,90), so spreadsheets separate columns with a semicolon. A semicolon CSV opened as if it used commas ends up crammed into a single column. 2) A comma inside a field, like "Smith, Ana", creates an extra column. The fix is to wrap the field in quotes: "Smith, Ana".
3) The encoding. If "José" shows up as "José", the file is UTF-8 and the program read it as another encoding. In Excel, import the original file from Data > From Text/CSV and select UTF-8.
If you paste the whole CSV into the chat, its rows use up context as tokens, and a long one costs money. Before pasting it, remove the columns you do not need. If the file is huge, ask the AI to help with a formula or an analysis instead of reading all 10,000 rows, and check the important totals yourself.
It works the other way too: you can ask it to give data back as CSV to paste into your spreadsheet. If you always need the same columns, a structured output or a script that fixes the format is better.