Safe CSV exports
A CSV value may be interpreted differently by a spreadsheet. Gridberry provides a Windows project workspace for discussing export code, reviewing escaping changes and checking adversarial cells before sharing a file.
function csvCell(value: string): string {
const guarded = /^(?:\s*[=+@-]|[\t\r\n])/.test(value) ? "'" + value : value;
return '"' + guarded.replaceAll('"', '""') + '"';
}Project
Check leading formula characters, tabs, commas and embedded quotes, while keeping the original application values unchanged.
Review before applying
Compare actual behavior with the expected result. If a check fails, investigate it before applying more changes.
Connection
Base URL, credentials and a reachable server.
function csvCell(value: string): string {
const guarded = /^(?:\s*[=+@-]|[\t\r\n])/.test(value) ? "'" + value : value;
return '"' + guarded.replaceAll('"', '""') + '"';
}Treat an export as a software contract
Gridberry is Windows AI coding software for working on the boundary between application data and the files people receive. Open the project folder and describe the export's intended reader, accepted values and expected representation. With the source and conversation together, you can ask a connected model to trace how data moves from a record to a downloadable file before it proposes an edit.
The csv-cell.ts example provides one concrete assignment: review how spreadsheet-bound values are guarded and quoted. Use it as a discussion starter for your own export requirements, not as an included CSV service or a guarantee about every spreadsheet program. A useful proposal explains both how ordinary text survives the conversion and how unusual leading characters are handled. Keep those decisions visible in the code review.
You may select an unrestricted model through an available supported OpenAI-compatible or Anthropic endpoint. Model restrictions, context and tools are determined by that connection, with credentials supplied when required. After reviewing an export patch, run the project's checks and open representative output in the actual consuming application. That final step matters because valid text and useful spreadsheet behavior are related but different outcomes. Preserve the input examples that motivated the change for later regression checks.
Checking data that leaves the application
Which sample values help review a spreadsheet export?
Include ordinary text, quotes, separators, line breaks and values with unusual leading characters or whitespace. State whether the reader should see text or a number. The expected interpretation is part of the requirement, not something quoting alone establishes.
Does quoting a CSV field settle spreadsheet formula handling?
No. CSV structure and a spreadsheet's interpretation are separate concerns. Review the intended guard behavior and test it in the programs your users rely on. Do not generalize one successful sample into a claim about every spreadsheet reader.
Can I keep exact text while changing export guards?
Decide which transformations your application accepts and document them in the task. Compare original values with the opened export, especially around prefixes and line breaks. Ask the model to identify tradeoffs instead of silently changing user-visible data.
What if the export patch affects other downloads?
Inspect shared serialization helpers and their callers before accepting it. Test affected formats individually. A helper change that fixes the spreadsheet path may alter an integration expecting the previous text, so make that compatibility decision explicit.
Gridberry · Updated 2026-10-25