Guide · Agent skill
How to use the Polars DataFrame Skill agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Polars DataFrame Skill is an agent skill that teaches an AI agent how to write and check Polars code for working with tables of data in Python. It is aimed at data analysts and engineers building ETL or reporting pipelines, and at developers moving pandas code to Polars. After this guide you will know how to install the skill, how to ask for Polars help in plain language, and what the skill does and does not do.
Polars DataFrame SkillK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the Polars skill ZIP from the directory page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use or .claude/skills/ for one project.
- 3Or upload in the apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Install Polars. Run uv pip install "polars==1.44.2" in a Python 3.10+ environment.
- 5Ask in plain language. Describe your table task and mention Polars or pandas, naming your files and columns.
- 6Test on a sample. Run the generated code on a small slice of your data and check the schema before using it on the full dataset.
What is an agent skill, and what does the Polars skill make the agent do?
An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches it. The Polars skill is defined by a SKILL.md file and teaches the agent Polars' expression-based syntax, lazy query optimization, and common patterns such as filtering, grouping, joins, reading and writing CSV, Parquet and JSON files, and moving code over from pandas.
When the skill is active, the agent can write Polars code for you and check it, and it can open reference guides for details such as I/O, best practices, core concepts, transformations, operations, pandas migration and review notes.
Install the Polars skill from a ZIP file
Download the ZIP for the skill from the directory page, then unzip it into one of these locations:
- Your personal skills folder: ~/.claude/skills/
- A single project's skills folder: .claude/skills/ inside the project
- Or upload the skill in the Claude apps under Settings → Capabilities → Skills
Confirm the skill folder is in place
After unzipping, the skill folder should contain SKILL.md and a references folder. The references folder holds best_practices.md, core_concepts.md, io_guide.md, operations.md, pandas_migration.md, review.md and transformations.md. The agent reads these files when it needs more detail.
Set up Polars in your Python environment
The skill requires Python 3.10 or newer. It uses Polars 1.44.2, which the skill says is installed with uv pip install. Optional extras cover Excel, database, fsspec, pandas and NumPy integrations, and you only need them if your work uses those features.
- Core install, as given in the skill
- Optional extras, only when needed
uv pip install "polars==1.44.2"
uv pip install "polars[excel,database,fsspec,pandas,numpy]==1.44.2"Trigger the skill with natural requests
You do not need a special command. Describe the table task in plain language and mention Polars or pandas so the agent knows which library you want. Good requests name the file type, the columns and the result you want.
Examples of requests that fit the skill:
- Read sales.csv with Polars, keep only rows where region is EU, and total revenue by month.
- Convert this pandas script to Polars and explain each change.
- Scan the large Parquet file lazily, filter to 2025 orders, and write the result to a new Parquet file.
- Join customers.json to orders.csv on customer_id using a left join.
- Review this Polars code for performance problems and suggest lazy mode where it helps.
Lazy versus eager: what the agent will recommend
Eager DataFrame code runs each step immediately. Lazy LazyFrame code builds a query plan that Polars optimizes before running. The skill says lazy mode is useful for large datasets, complex pipelines, cases where only some columns or rows are needed, and performance-critical work. Its benefits include predicate pushdown, projection pushdown and parallel execution.
lf = pl.scan_csv("file.csv")
result = lf.filter(pl.col("age") > 25).select("name", "age")
df = result.collect()Safety notes: scripts and network
The skill's listed requirements say it does not run scripts and does not use the network. It lists Read as its allowed tool. Cloud, GPU, BigQuery and remote database recipes in the references are described as illustrative and need provider setup, so test them before relying on them in production pipelines. Always check any output path the agent chooses before writing files.
- Runs scripts: no
- Uses network: no
- Cloud, GPU, BigQuery and remote database recipes need your own provider setup and testing
Tips for getting better results
Give the agent the real column names and file paths, because the skill's fragments use illustrative names that must be adapted to your dataset. Ask for the schema to be validated after pandas conversions, since Polars has no index and types are inferred. Run the generated code on a sample first.
- Paste a few sample rows or the schema.
- State the Polars version you use, ideally 1.44.2.
- Ask the agent to load the matching reference file for I/O or migration questions.
What people use it for
- Converting a pandas cleaning script to Polars expressions.
- Scanning a large CSV or Parquet file lazily and writing a filtered result.
- Building group-by summaries and window statistics such as group averages with over().
- Joining and concatenating tables from CSV, Parquet or JSON sources.
- Reviewing existing Polars code for slow patterns and suggesting lazy mode.
Frequently asked questions
Is the Polars agent skill free?
The skill is published under the MIT licence, as listed on the directory page. Polars itself is a library you install with uv pip install, and the skill's license field points to the Polars LICENSE file on GitHub.
What do I need to use the Polars skill?
You need Python 3.10 or newer and Polars 1.44.2 installed with uv pip install. Optional extras are only needed for Excel, database, fsspec, pandas and NumPy work, and cloud, GPU or BigQuery recipes need their own setup.
Where do I put the skill folder?
Unzip the downloaded skill into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
Does the Polars skill run code or access the internet?
The listed requirements say it runs no scripts and uses no network access. Its allowed tool is Read. Cloud and database recipes are illustrative and need provider setup.
Can the skill help me move from pandas to Polars?
Yes. The skill covers pandas migration, including a mapping table for common operations such as selecting columns and filtering, and it notes that Polars has no index and uses typed columns.
What is the alternative if I only need a spreadsheet?
This skill is not meant for non-programmers who only need a spreadsheet, because it assumes Python and command-line installs. If you are not writing Python, a spreadsheet tool is the better fit.
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
Use the Polars skill if you write Python for ETL, reporting or pandas-to-Polars migration and want an agent that writes and checks Polars code with reference guides behind it. Skip it if you only need spreadsheet work or if you cannot install Python packages, and test any cloud, GPU or database recipes before using them in production.
Download the skill