How to use the Dummy Dataset Generator agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent reads when a task matches the skill's description. This skill is named dummy-dataset and its description says it generates realistic dummy datasets with customizable columns, constraints and output formats.

When it is installed, you describe the kind of data you need, such as customer feedback records or transaction lists, along with the columns, row count and any business rules. The agent then follows the skill's step-by-step process: identify the dataset type, define column specifications, pick a row count, choose an output format, apply realistic patterns, add business constraints, generate the data and validate it.

• Output as CSV, JSON, or SQL INSERT statements
• Output as a Python script that generates the data
• Rules such as skewed rating distributions or category-to-rating limits

Install the dummy-dataset skill

Download the ZIP for the skill from the pm-execution/skills/dummy-dataset page on our directory. Then choose one of the following methods.

For a personal install that works across your projects, unzip the download into ~/.claude/skills/. For a single project, unzip it into the .claude/skills/ folder inside that project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

• Personal: unzip into ~/.claude/skills/
• Single project: unzip into .claude/skills/ in the project folder
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Ask for a dataset with natural requests

You do not need a special command. The skill is meant to be triggered when you ask for test data, mock datasets or sample data for development or demos. Describe the dataset type, the columns, the number of rows, the output format and any constraints in one message.

If you leave something out, the skill's defaults and steps still apply. The skill's row count default is 100 rows, and the agent is expected to ask about or infer the other details from your request.

• "Create 500 rows of customer feedback as CSV with id, name, email, date, rating and category."
• "Make a SQL INSERT file with 200 fake transactions for my staging database."
• "Generate a Python script that builds 10,000 user profiles and saves them as CSV."

Example: a customer feedback dataset with rules

The skill's example specification shows how constraints work. It defines a customer feedback dataset with columns such as feedback_id, customer_name, email, feedback_date, rating, category, text and product. Its constraints include a rating skew of 40% 5-star, 30% 4-star, 20% 3-star and 10% 1-2 star, and rules that Bug entries only carry ratings 1-3 while Feature Requests only carry ratings 3-5.

You can write your own rules in the same way. The more specific the rule, the easier it is to check the output.

Choose an output format

CSV is a flat table that opens in spreadsheets and imports into databases. JSON has a structure suited to APIs and NoSQL databases. SQL gives INSERT statements that run directly on relational databases. The Python script option produces a generator you can edit and run yourself, which suits large or custom datasets.

The skill's Python template writes a CSV file and prints how many records it generated. Its column definitions are placeholders that the agent fills in for your dataset.

• CSV: spreadsheets and database imports
• JSON: APIs and NoSQL databases
• SQL: INSERT statements for relational databases
• Python script: editable generator for custom or large datasets

Safety notes: scripts and network access

This skill runs scripts, so review any Python the agent writes before you run it, especially on a machine with access to important files. The skill does not use the network, according to its listing. Python is needed only if you choose the Python script output or run the generated script.

The data is invented. Do not use this skill to protect or anonymise real personal data, because it creates new fake records rather than masking existing ones. Do not use the output to analyse real customer behaviour, because the results will not reflect real people.

• Review generated scripts before running them
• Python is required only for the script output or running that script
• No network use is listed for this skill
• Not a tool for masking or anonymising existing records

Tips for better output

Name the exact columns and their formats, such as dates in the last 90 days or emails from gmail, yahoo and company.com. State the row count and any distribution you need. Ask the agent to validate the output, since the skill's process includes a validation step for data quality and constraint compliance.

Check a sample of rows against your rules before loading the data into a database or a demo.

• Give exact column names and value ranges
• State business rules explicitly
• Ask for validation against your constraints
• Inspect a few rows before importing