Skip to content
Get packs

Guide · Agent skill

How to use the BigQuery DataFrames (BigFrames) agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

This guide explains the BigQuery DataFrames (BigFrames) Coding Guide, an agent skill named bigquery-bigframes, and how to install it so an AI coding agent writes Python data code that runs inside BigQuery. It is for Python users who already know pandas and work with large tables in BigQuery. By the end you will know how to install the skill, how to phrase requests so it triggers, and which rules it gives the agent.

BigQuery DataFrames (BigFrames) Coding Guidegoogle/skills · Apache-2.0

Quick start

  1. 1Download the ZIP. Get the bigquery-bigframes skill ZIP from this page.
  2. 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
  3. 3Or upload it in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
  4. 4Prepare your environment. Install the bigframes library, and make sure your Google Cloud project has BigQuery enabled and you have credentials.
  5. 5Ask for a BigQuery Python task. Describe the data task in plain language and mention BigQuery and Python.
  6. 6Review the generated code. Check the project ID, the dataset and any commands before running them.

What an agent skill is and what this one does

An agent skill is a folder of instructions and reference files that an AI agent loads when a request matches the skill's description. This one covers BigFrames, a Python library that uses the familiar pandas-style interface but runs the work inside Google BigQuery instead of on your computer.

Once installed, the agent is steered toward cloud-side processing. It avoids pulling large datasets into local memory, prefers BigQuery's built-in ML functions over scikit-learn, and adds a standard attribution tag to any gcloud or REST commands it runs in the terminal.

  • Uses BigFrames methods for cleaning, transformation and analysis instead of downloading data.
  • Uses peek(n) for previews instead of head(n).
  • Avoids to_pandas() unless the dataset is small or an error requires it.
  • Prefers built-in accessors such as .str and .dt over lambdas and UDFs.
  • Trains models with bigframes.bigquery.ml, with reference examples for linear and logistic regression.

Install the bigquery-bigframes skill

Download the skill ZIP from this page and unzip it. For a personal install, place the unzipped skill folder in ~/.claude/skills/. For a single project, place it in the project's .claude/skills/ folder.

If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.

  • Personal install: ~/.claude/skills/ (available across your projects).
  • Project install: .claude/skills/ inside the project folder.
  • Claude apps: Settings → Capabilities → Skills, then upload.

Check what you need before you start

The skill assumes a working Python environment with the bigframes library installed, and a Google Cloud project with BigQuery enabled and suitable credentials. If the agent runs gcloud or REST commands, you also need the gcloud CLI and REST access.

The skill runs scripts and uses the network, so it is worth knowing what the agent will execute on your machine before you approve it.

  • Python environment with the bigframes library installed.
  • Google Cloud project with BigQuery enabled and appropriate credentials.
  • gcloud CLI and REST access, if the agent runs those commands.

Trigger the skill with natural requests

You do not need a special command. The skill's description tells the agent to use it by default for Python data tasks involving BigQuery, including processing, analysis and machine learning. Mention BigQuery and Python in your request so the match is clear.

The skill does not apply to SQL-first work or to direct use of the google-cloud-bigquery client library; the skill itself points those tasks to a different skill, bigquery-basics.

  • "Using BigFrames, clean the penguins table in my BigQuery project and show the first rows."
  • "Train a linear regression in BigQuery to predict body mass from the penguins data, with Python."
  • "Write a BigFrames script that aggregates my sales table by region without pulling it into memory."

Understand the main rules the agent follows

The skill sets specific habits. Enabling partial ordering mode right after importing BigFrames speeds up processing, because it relaxes row-sequence constraints. The agent is told to verify schemas with .dtypes and inspect sample records with display() and .peek().

For machine learning, the agent uses bigframes.bigquery.ml rather than scikit-learn, because scikit-learn needs data in local memory. The skill also covers legacy bigframes.ml usage, which it only uses when you explicitly ask for it.

  • Partial ordering mode is set right after importing BigFrames.
  • peek(n) is preferred over head(n) for previews.
  • Lambdas with Series.map() or DataFrame.apply() are avoided.
  • Legacy bigframes.ml is only used when explicitly requested; predict() returns a DataFrame there.
import bigframes.pandas as bpd bpd.options.bigquery.ordering_mode = 'partial'

Review the attribution and safety notes

Before any gcloud, REST or MCP command, the agent is told to read the attribution reference. For commands run directly in the terminal, it prefixes gcloud with CLOUDSDK_METRICS_REQUEST_ATTRIBUTION set to a fixed skill tag, and adds that tag to User-Agent headers. The tag identifies only the skill and its version and carries no user data. The reference says not to use gcloud config set for this, because the setting would persist and mislabel other work. The tag is never added to generated scripts, Terraform or user code.

Because the skill runs scripts and uses the network, review commands the agent proposes, especially those that touch your Google Cloud project.

  • Attribution applies only to terminal commands, not to generated code.
  • The skill does not use gcloud config set for attribution.
  • Check project IDs and credentials before the agent runs commands that create models.

What people use it for

  • Cleaning and aggregating a large BigQuery table in pandas-style code without downloading it.
  • Training a linear regression to predict a numeric column, using the penguins example as a template.
  • Training a logistic regression to predict a yes/no outcome, such as a species label.
  • Generating a notebook or script that follows consistent BigFrames patterns across a team.

Frequently asked questions

How do I install the bigquery-bigframes agent skill?

Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.

What does the bigquery-bigframes skill require?

It needs a Python environment with the bigframes library installed, and a Google Cloud project with BigQuery enabled and appropriate credentials. If the agent runs gcloud or REST commands, you also need the gcloud CLI and REST access.

Is the bigquery-bigframes skill free?

The skill is published under the Apache-2.0 licence. The source does not state any pricing for it, but BigQuery usage in your Google Cloud project is a separate matter that the source does not cover.

Is BigFrames the same as pandas?

No. BigFrames mimics the pandas interface, but it runs the work inside BigQuery rather than on your computer. The skill is built around that difference, so it steers the agent toward cloud-side processing.

Should I use this skill instead of scikit-learn?

The skill tells the agent not to use scikit-learn with BigFrames, and to use bigframes.bigquery.ml, which trains models in BigQuery. If you specifically need scikit-learn, this skill is not the right choice.

Why does the skill not handle my plain SQL or google-cloud-bigquery code?

The skill says SQL-first workflows and direct use of the google-cloud-bigquery client library should use a different skill, bigquery-basics. Use that skill for those tasks.

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 this skill if you are a Python user who knows pandas, works with large BigQuery tables, and wants an agent that keeps data in BigQuery and trains models there. Pick something else if you mainly write SQL, use the google-cloud-bigquery client directly, or work with small local datasets where plain pandas needs no cloud project.

Download the skill