kaggle-skill plugin
kaggle-skill answers Kaggle competition questions with sources, tracks your standing and submits only after you agree.
- 112 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skill submits solutions, runs notebooks and publishes datasets from your account after confirmation and spends daily limits and GPU hours.
- The Kaggle token is stored on disk in `~/.kaggle`.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/kaggle-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill submits solutions, runs notebooks and publishes datasets from your account after confirmation and spends daily limits and GPU hours.
- The Kaggle token is stored on disk in `~/.kaggle`.
This is third-party code. Review the repository files before installing.
What it does
An unofficial skill and plugin named `kaggle` for Claude Code, Codex and other agents. It answers questions about competitions: metric, deadline, prize, rules, winner writeups and discussions. With your credential it shows rank, submissions left today, best score and GPU hours. Actions with consequences go through a dry run: submitting a solution, running a notebook, publishing a dataset or model needs your consent with the `--yes` flag. The skill checks a submission file, downloads data and reads public pages on its own. Text from Kaggle is marked as data so an instruction hidden in a forum post is not followed.
Who it is for. Kaggle participants who want to run competitions through an agent: tracking standing, checking files and studying other solutions.
Good fit when
- You need the metric, deadline and prize of a competition from a link.
- A submission file format must be checked before sending.
- You want a summary of what top teams did.
- A notebook runs on Kaggle and the score is awaited.
Not a fit when
- You have no Kaggle account or token for your own data.
- You need an official tool from Kaggle only.
- The agent must never submit solutions even after confirmation.
Example request
Which Kaggle competitions with prizes are running now? Give a short list with deadline, prize and metric.Limitations
The project is independent and not affiliated with Kaggle or Google. Python 3.11 is required, and account work needs the `kaggle` package 2.2.4 and `kagglehub` 1.0.2. Sign in with `kaggle auth login` or a token from settings. OAuth login does not work for `kagglehub`, so publish with `--via cli`. `KAGGLE_SKILL_READ_ONLY=1` blocks every write. Local records of submissions live in `./.kaggle-skill/`.
How to disable. Remove the plugin with `/plugin uninstall kaggle@shepsci` or the `skills/kaggle` folder in the agent directory. Revoke the Kaggle token in account settings and clean `./.kaggle-skill/` and `~/.kaggle` by hand.
FAQ
How do I install it in Claude Code?
Run `/plugin marketplace add shepsci/kaggle-skill`, then `/plugin install kaggle@shepsci`.
How do I install it in Codex?
Run `codex plugin marketplace add shepsci/kaggle-skill --ref main` and `codex plugin add kaggle@shepsci`.
What does it do without asking?
It reads public pages and your data, downloads competition data and checks a submission file. Submitting a solution or running a notebook needs your consent.
How do I block all writes?
Set `KAGGLE_SKILL_READ_ONLY=1`.
How do I check the setup?
Ask "check my Kaggle setup" and the skill runs its `doctor` command.
Discussion
No questions or reviews yet.
Related
PluginsA skill that runs a Kaggle competition from rules and folds to a scored submission, with hidden-rerun debugging and ensembles
The NVIDIA plugin guides an agent through a Kaggle competition: rules, winner writeups, public notebooks and solution submission.
For a repeatable experiment, an agent downloads a Kaggle dataset to a working directory and retrieves notebook outputs.
A template for data scientists: AGENTS.md, CLAUDE.md and skills for Python, SQL, Polars and model diagnostics.
31 skills for Claude Code and Codex for the analyst workflow: data checks, cohorts, funnels, A/B tests, dashboards, summaries
Probabl skills for machine learning experiments: a scikit-learn and skrub pipeline, evaluation, audit and a link to Skore Hub
README in short
The README shows a real session, requirements, installation for Claude Code, Codex, skills.sh and OpenClaw, sample requests and a table of actions the skill does itself or after asking. A diagram of pieces and sign-ins is included.