Guide · Agent skill
How to use the Data Cleaning Specialist agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Data Cleaning Specialist is an agent skill that teaches an AI agent how to clean messy spreadsheets and exported tables. It checks the data for problems, agrees cleaning rules with you, and delivers a new Excel workbook with a standardised table and a log of every change. This guide is for analysts and operations staff who receive data from several systems, and it explains how to install the skill, trigger it with plain requests, and review its output safely.
Data Cleaning Specialistgenspark-ai/genoffice · Apache-2.0 Download the skill Quick start
- 1Install the skill. Unzip the downloaded ZIP into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 2Confirm spreadsheet tools are available. Check that the agent has spreadsheet tools, since the skill needs them to build the Excel workbook.
- 3Make a clear request. Describe the data and goal, for example asking the agent to clean a customer list and remove duplicates.
- 4Answer the downstream-use question. Tell the agent whether the result is for a system import, analysis or mailing, because that decides the standard format.
- 5Confirm the cleaning rules. Review the proposed rules for blanks, duplicate keys and formats. Approve or change them before any deletion runs.
- 6Review the workbook. Open the Cleaning log and Suspected duplicates sheets to check every change against the Raw data sheet.
- 7Decide the flagged rows. Work through the row numbers the agent could not decide automatically.
What the data-cleaning skill makes the agent do
An agent skill is a folder with a SKILL.md file that gives the agent a set of instructions for one kind of task. Once installed, this skill makes the agent act as a data-cleaning specialist for system exports, merged tables and manually keyed sheets.
The agent works in a fixed order. It first runs a health check that covers blanks, mixed types such as text inside number columns or several date formats, obvious duplicates, inconsistent units and codes, and stray spaces or full-width versus half-width characters. It then asks what the data is for, since that decides the standard format, and proposes rules for each problem type.
- Health check: lists field meanings, row count, blank rate per column and detected problems.
- Rules: one rule per problem class, such as how to handle blanks, which columns define a duplicate, and date and number formats.
- Confirmation: the rules are sent to you first, and anything involving deletion must be confirmed before it runs.
- Output: a new workbook in which the original data is left untouched.
Install the Data Cleaning skill
The skill is distributed as a ZIP download from the directory page, and its only listed file is SKILL.md. You can install it in one of two ways, depending on where you use the agent.
For Claude Code, unzip the download into your personal skills folder ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it available only in that project. For the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/ so the skill is available everywhere.
- Project install: unzip into the project's .claude/skills/ folder.
- Claude apps: upload the ZIP under Settings → Capabilities → Skills.
Check the requirements before you start
The skill needs spreadsheet tools available to the agent, because it builds the Excel workbook with them. If the agent has no spreadsheet tools, it cannot produce the deliverable.
The skill runs no scripts and uses no network access, according to its listing. It works on data you provide in the conversation or workspace.
- Requires: spreadsheet tools available to the agent for building the Excel workbook.
- Runs scripts: no.
- Uses network: no.
- Best suited to small and medium tables, since the workflow relies on the agent's spreadsheet tools and step-by-step confirmation.
Trigger the skill with natural requests
You do not need a special command. Describe the cleaning job in plain language and mention the data. The skill's suggested prompts are a good starting pattern.
Good requests name the file or table, the goal and the expected output. Mention if you need one standard layout for several sources.
- Clean this customer list and remove duplicates.
- Merge these inconsistently formatted tables into one standard table.
- Find blanks, mixed date formats and stray spaces in this export and propose rules before changing anything.
- Standardise these three supplier sheets into one layout and log every change.
Review the output workbook
The result is one Excel workbook titled "<Dataset> Cleaning Result" with five sheets. Each sheet has a clear job, so you can check the work instead of trusting the final table.
The Cleaning log is the most important sheet for an audit. It records the original row, field, old value, new value, rule number and action (modify, delete or merge). When you merge several tables, the Cleaned sheet keeps a source-table column so you can trace each row back.
- Raw data: your original table as supplied.
- Cleaned: the standardised result with unified headers, one type per column and ISO dates.
- Cleaning rules: rule number, field, problem, rule and rows affected.
- Cleaning log: every modified, deleted or merged record.
- Suspected duplicates: groups that may be duplicates, with the kept record marked for your review.
Handle records that need a human decision
The skill does not merge records on its own when names look similar but keys differ. Those cases go to the Suspected duplicates sheet. Place or company names that look misspelt are only changed when a mapping table supports the change; otherwise they are marked as suspected and left alone.
After delivery, the agent summarises the row counts from original to final, listing deleted, merged and modified records. It also lists the row numbers it could not decide automatically.
- Similar names with different keys are left for you to decide.
- Units are never assumed; the agent asks for the basis before converting.
- National ID numbers, phone numbers and e-mail addresses are only format-checked, never inferred or completed.
What people use it for
- Combining customer exports from a CRM and a mailing tool into one contact table with a consistent layout.
- Standardising dates, numbers and codes across monthly operations sheets before analysis.
- Removing duplicate records from a manually keyed list, with suspected duplicates reviewed by a person.
- Merging several supplier spreadsheets into one table with a source-table column for traceability.
Frequently asked questions
How do I install the Data Cleaning Specialist agent skill?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
What does the data-cleaning skill require?
It requires spreadsheet tools available to the agent, because it builds the Excel workbook with them. According to its listing, it runs no scripts and uses no network access.
Will the skill delete my data without asking?
No. The skill is designed to propose rules and ask for confirmation before any deletion runs. Every deleted or merged record is also written to the cleaning log.
Does the skill change my original file?
No. The original data is kept in the Raw data sheet, and the cleaned result is produced in new sheets of a separate workbook.
Can it clean sensitive data like phone numbers and emails?
It only format-checks and standardises fields such as national ID numbers, phone numbers and e-mail addresses. It never infers or completes them, but you should still handle the output file carefully because it is a spreadsheet that may be distributed.
Is it suitable for very large datasets?
Probably not. The workflow relies on the agent's spreadsheet tools and step-by-step confirmation, so it is best for small and medium tables. For very large datasets, consider a dedicated data tool.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Choose this skill if you regularly merge inconsistent exports and need one standard table with an auditable record of every change. Pick something else if you need fully automatic deletion without review, you are working with very large datasets, or you do not have spreadsheet tools available to the agent.
Download the skill