Skill library / DeepSeek

Spreadsheet Data Cleaner skill for DeepSeek

Spreadsheets pick up errors fast: names in mixed case, dates in three formats, stray spaces, duplicate rows. Fixing them by hand takes hours and is easy to get wrong. This skill has DeepSeek inspect pasted data, propose consistent rules, apply them, and list exactly what changed so you can trust the result.

What it's good for

• Cleaning a customer mailing list before a mail merge, fixing capitalisation, postcodes and duplicate entries
• Standardising dates and amounts in a monthly expenses export so they sort and total correctly
• Tidying a supplier contact sheet merged from several people's files, with mixed phone number formats and blank rows

How the skill works

1. Ask the user to paste the data (or a representative sample of up to about 50 rows) with the header row included, and ask what the cleaned sheet will be used for.
2. Scan the data and list the problems found by column, for example mixed date formats, trailing spaces, inconsistent capitalisation, or likely duplicates, with a count and one example of each.
3. Propose a clear rule for each problem, such as UK dates as DD/MM/YYYY, phone numbers in a single format, names in title case, and ask the user to confirm or change them before you proceed.
4. Apply the agreed rules and return the cleaned data as a table or comma-separated text the user can paste straight back into their spreadsheet, keeping the original column order.
5. Finish with a change log listing each rule applied and how many rows it affected, then list any rows you were unsure about, flagged separately and left unchanged for the user to decide.

Built-in rules

• Never delete rows, including apparent duplicates, without showing them to the user first and getting agreement.
• Do not guess missing or ambiguous values, such as a date like 03/04/2024 or a blank postcode. Flag them instead.
• Keep the original wording of free-text columns such as notes or comments unless the user asks you to edit them.

Example

You: “Here are 40 rows from our customer list. Names are all over the place, dates are mixed up and I think there are duplicates. It's going into a mail merge.”
Result: A cleaned table with names in title case, dates as DD/MM/YYYY and consistent postcodes, plus a change log and a short list of 3 suspected duplicates and 2 ambiguous dates awaiting your decision.

SKILL.md template

Save as spreadsheet-data-cleaner/SKILL.md.

---
name: spreadsheet-data-cleaner
description: "Tidies messy spreadsheet data you paste in, such as inconsistent names, dates, phone numbers and duplicates, and explains each change. Use it when a customer list, expenses sheet or export is a mess and you need it ready for sorting, mail merge or import."
---

# Spreadsheet Data Cleaner

## Workflow
1. Ask the user to paste the data (or a representative sample of up to about 50 rows) with the header row included, and ask what the cleaned sheet will be used for.
2. Scan the data and list the problems found by column, for example mixed date formats, trailing spaces, inconsistent capitalisation, or likely duplicates, with a count and one example of each.
3. Propose a clear rule for each problem, such as UK dates as DD/MM/YYYY, phone numbers in a single format, names in title case, and ask the user to confirm or change them before you proceed.
4. Apply the agreed rules and return the cleaned data as a table or comma-separated text the user can paste straight back into their spreadsheet, keeping the original column order.
5. Finish with a change log listing each rule applied and how many rows it affected, then list any rows you were unsure about, flagged separately and left unchanged for the user to decide.

## Rules
- Never delete rows, including apparent duplicates, without showing them to the user first and getting agreement.
- Do not guess missing or ambiguous values, such as a date like 03/04/2024 or a blank postcode. Flag them instead.
- Keep the original wording of free-text columns such as notes or comments unless the user asks you to edit them.

How to install in DeepSeek

Copy the template above, then open DeepSeek. Start a new chat, paste it in and add: "Please follow these instructions for this chat." DeepSeek can't save instructions yet, so keep our email handy and paste it again next time.

Pro tip

DeepSeek does not remember earlier chats, so paste the skill instructions at the start of each new session, then the data. For large sheets, clean them in batches of 50 to 100 rows and reuse the same agreed rules each time so the results stay consistent.

FAQ

Related skills

• Revision Timetable skill for DeepSeek
• Plain English Explainer skill for DeepSeek