Katas First steps · About 10 minutes

Rescue messy data

Turn a small, messy spreadsheet into a useful report you can check.

Start this kata See the result

01 / See it first

Here’s what you can make

Example result · synthetic data

A report you can trust

4 complete sales

2 need review

$22 complete sales total

Seven rows came in. Remove one exact duplicate, then set aside the two rows with missing or invalid values. The four complete sales contain seven items: six Basil for $18 and one Mint for $4. Try the checks by hand or use the assistant prompt below.

Download the practice CSV

This is fictional practice data created for the kata. The figures below are calculated from the dataset, not real sales.

02 / Your turn

Copy the prompt. Make it yours.

Open a new chat in ChatGPT or Claude, paste this prompt, and replace anything you want to change.

Your complete prompt

You are my data-quality partner. Clean this fictional seed-sale dataset and write a short, honest report.

Rules:
- Treat exact duplicate rows as one sale. Treat item names without regard to capitalization.
- Do not guess missing quantities or non-numeric prices. Put incomplete rows in an exceptions list.
- Calculate quantities and revenue only for complete, unique rows. Show your arithmetic.
- Report the number of rows received, exact duplicates removed, complete unique sales and exceptions.
- Include a short table of the clean rows and a short plain-language summary.
- This is synthetic practice data: do not describe the sales as real or invent missing values.

Here is the CSV:

id,date,item,quantity,unit_price_usd
S101,2026-04-01,Basil,2,3.00
S102,2026-04-01,Mint,1,4.00
S103,2026-04-02,Basil,3,3.00
S102,2026-04-01,Mint,1,4.00
S104,2026-04-02,Mint,,4.00
S105,2026-04-03,Basil,2,unknown
S106,2026-04-03,basil,1,3.00

You can also select and copy the text.

Make this prompt about your idea

These details stay in this page. The Dojo does not save or send them; you choose whether to copy and share the result.

What do you want to make?How much time do you have? (optional)What tools can you use? (optional)What else should it know? (optional)

03 / Your ingredients

A simple stack

Harness: ChatGPT or Claude app, or either service in a browser.
Model: A model available in your chosen chat. Access and limits depend on your account.
Instructions: The complete prompt above. Change the topic and constraints to fit your goal.
Context: Your idea and any non-sensitive details you choose to share.
Skills, CLI, MCP, plugins or connectors: None needed for the basic chat version. Optional ways to equip your setup are below.
Agent and verification: One chat assistant drafts; you inspect the result against the checks below and ask it to fix problems.

04 / Pick your setup

Use the device you have

Choose one path. You can do the chat exercise without installing a command line tool.

Phone · ChatGPT app

Open the ChatGPT app, start a new chat, and paste the prompt. Ask the assistant for the requested result. Some account-available plugins can be used on mobile; desktop-only plugins cannot. Plugins are optional here. ChatGPT plugin guide.

Phone · Claude app

Open the Claude app, start a new chat, and paste the prompt. Ask for the requested result. Remote connectors are optional; local desktop extensions do not run on the phone. Claude connectors guide.

Windows

Open ChatGPT or Claude in a browser and paste the prompt. Alternatively, install one CLI in PowerShell using an official command below. These commands download and execute installers; check the linked vendor instructions first. Sign in when prompted; eligibility and plans vary. Use a folder you are comfortable granting the tool access to.

Codex CLI (official setup):

powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"
codex

Claude Code (official setup; an eligible paid or other supported account is required):

irm https://claude.ai/install.ps1 | iex
claude

Then paste the same prompt into the CLI. Review any access or changes it asks to make.

macOS

Open ChatGPT or Claude in a browser and paste the prompt. Alternatively, install one CLI in Terminal using an official command below. These commands download and execute installers; check the linked vendor instructions first. Sign in when prompted; eligibility and plans vary. Use a folder you are comfortable granting the tool access to.

Codex CLI (official setup):

curl -fsSL https://chatgpt.com/codex/install.sh | sh
codex

Claude Code (official setup; an eligible paid or other supported account is required):

curl -fsSL https://claude.ai/install.sh | bash
claude

Then paste the same prompt into the CLI. Review any access or changes it asks to make.

Linux

Open ChatGPT or Claude in a browser and paste the prompt. Alternatively, install one CLI in a terminal using an official command below. These commands download and execute installers; check the linked vendor instructions first. Sign in when prompted; eligibility and plans vary. Use a folder you are comfortable granting the tool access to.

Codex CLI (official setup):

curl -fsSL https://chatgpt.com/codex/install.sh | sh
codex

Claude Code (official setup; an eligible paid or other supported account is required):

curl -fsSL https://claude.ai/install.sh | bash
claude

Then paste the same prompt into the CLI. Review any access or changes it asks to make.

05 / Check and remix

Make sure it works for you

Check that the repeated S102 row is counted once, and that S104 and S105 are listed as exceptions.
Check the complete rows: 2 × $3, 1 × $4, 3 × $3 and 1 × $3. The total is $22 across seven items.
Check that “Basil” and “basil” are grouped together and that missing values were not invented.
If a check fails, tell your assistant exactly which row is wrong and ask for a corrected report.

Check your three numbers

This checks the numbers for the fictional seed dataset. You still need to inspect the rows and the explanation.

Complete unique salesRows needing reviewComplete sales total in dollars

Remix: use a small dataset you are allowed to share, and define how to handle duplicates and missing values before calculating.

Explore more katas