AI coding mission · CSV cleaner

Generate a Python CSV cleaner with tests

Turn a description of your messy CSV into a Python cleaning script you can inspect and run locally. PlushCrew’s coding agents plan the work, write the files and attempt to test them in a sandbox. You get the project files and test output, so you can check the cleaning rules before using them on your data.

Try this mission free

Fills the task box; you choose when to send. Guest access: 1 mission per IP in a rolling 24 hours.

Describe the result you need

Useful CSV cleaning starts with precise rules. Include the input column names, a few fictional rows, the desired date convention and whether blanks should stay blank. Specify which columns define a duplicate: comparing an entire row is different from comparing only a customer ID. If you do not know a rule yet, use Chat to discuss it before spending a Mission.

Example prompt

Create a Python CSV cleaner that removes duplicate rows, converts dates to YYYY-MM-DD, and formats decimal numbers consistently. Preserve the header and row order, report invalid values, and add tests using a small example CSV.

Example project files and output

The exact files depend on your request. A project for this prompt could include:

clean_csv.pyThe cleaning function and a command-line entry point.
tests/test_clean_csv.pyTests for duplicate handling, date parsing and numeric formatting.
README.mdInstructions for running the script against your own CSV.

name,date,total
Ada,2026-10-01,1234.50
Linus,2026-10-02,99.00

Illustrative cleaned output: one consistent date format, decimal format and row per input record after deduplication.

What to check in the tests

Review the test output in the mission report and open the generated test files. Useful checks for this task include:

Duplicates are removed without changing the first occurrence’s position.
Quoted commas, empty cells and invalid dates have explicit expected behavior.
The script writes to a separate output file and keeps the original input intact.

Passing tests show what the supplied cases covered. They do not guarantee that every input will work or that generated code is safe.

Inputs and limitations

Dates such as 03/04/2026 are ambiguous without a locale. Tell the agents whether this means March 4 or April 3. Likewise, a comma can be a decimal separator or a thousands separator. Review those decisions in the code and tests. The mission creates a script; it does not upload and clean your local CSV automatically.

Review first, then download

Use Mission for a new project. The existing agent pipeline, sandbox and logic checks run as usual. Open each file in the browser, expand the report to inspect test results, then download the ZIP to use the project locally. A dispatched guest mission consumes the free slot even if it fails or you cancel it.

Chat is for questions; guests have five messages per IP in a rolling 24-hour window. For fixing an existing codebase, Audit accepts a code ZIP from a Pro Max account. See plans and limits for account details.

Fill this example task