Claude Code: Build CSV Exports That Match Your Dashboard

Define the export as part of the feature

The starting mismatch is deliberate: the interface shows three tickets, but the export contains twelve. A useful brief names the active filters, required columns, and expected rows. It also says which parts of the app the change may touch. That turns a vague request for an export button into a behavior a reviewer can compare with the dashboard.

Check the data that actually leaves the app

Use the downloaded file as evidence. Parse it with an independent reader and compare the records and fields against the expected filtered result. CSV quoting matters when values contain characters that could be mistaken for separators or line breaks. A clean-looking table in the browser cannot establish that the exported file preserved those values.

Include dates and empty results

Filters must compose: a date choice should agree with the other active criteria. The episode checks inclusive dates and the state where no records match. Decide what the empty download should contain and what the interface should explain. These ordinary user choices reveal mismatches that a single successful export can miss.

Read the example’s limits

The support app, seeded failure, and working reference were authored for this tutorial. The broken starter is not attributed to Claude, and the episode is not a measured model benchmark. Its useful output is a reproducible feature brief and a way to inspect the result. Apply those checks to the implementation and data rules of your own app.

Jump to a chapter