Claude Skill

kaggle

Operate Kaggle competitions, datasets, kernels, and submissions.

LLM Mart · 0 points · 3 views 0 listing impressions 0 install-command copies
Download fmind-dot-skills_kaggle-3bbe830.zip · 1 KB

Install

skills CLI npx skills add https://github.com/fmind/dot/tree/main/skills/kaggle
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install fmind-dot@llmmart
Git git clone https://github.com/fmind/dot.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole fmind/dot collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Kaggle CLI

Operate Kaggle competitions, datasets, kernels, and models via kaggle (auth, download bounds, and authority gates on submissions, remote runs, and publications).

Workflow

  1. Verify account: kaggle auth login (OAuth), env KAGGLE_API_TOKEN, or legacy ~/.kaggle/kaggle.json; never run kaggle auth print-access-token.

  2. Inspect before downloading: check bounded listings, kaggle competitions submission-limits <slug> --json, and kaggle competitions pages list list <slug> --content --page-name rules (double list required by the parser; reading rules never authorizes accepting them).

  3. Download into gitignored data/ (accept competition rules in the browser first, else HTTP 403):

    kaggle competitions download <slug> -p data/ --unzip
    kaggle datasets download <owner>/<name> -p data/ --unzip
    
  4. Check deadlines: inspect kaggle competitions pages list list <slug> (--page-name timeline, else description) and offer—or schedule when asked—gws Calendar reminders one week before entry/merger deadlines and on final submission day.

  5. Audit kernels before push: kaggle kernels init -p <dir> creates kernel-metadata.json. Before authorized kaggle kernels push -p <dir>, verify upload files, target ID, data sources, is_private: true (default private), internet, --accelerator, -t timeout, --no-run (push executes immediately otherwise), and kaggle quota; check kernels status <owner>/<slug> and kernels output <owner>/<slug> -p out/.

  6. Authorize submissions (each counts against daily caps and posts to the leaderboard; --wait exits non-zero on scoring error/timeout after consuming the slot, or check competitions submission <ref>):

    kaggle competitions submit <slug> -f submission.csv -m "<message>" --wait 900
    kaggle competitions submissions <slug> --format json --page-size 5
    
  7. Authorize dataset publications: kaggle datasets create -p <dir> defaults to private unless --public is approved; datasets version -p <dir> -m "<message>" keeps existing visibility (verify first, confirm license/audience, and pass history-destroying -d/--delete-old-versions only when requested).

Gotchas

  • Prefer uv run kaggle when a project pins kaggle so the locked version runs over the global shim.
  • Strip import-time Warning: lines before JSON parsing (-W takes effect after import-time auth checks).

Official Skills

Upstream: Kaggle/kaggle-cli (CLI commands) and Kaggle/kaggle-skills (judging, exams, Kaggle Benchmarks), reviewed per vendor-skill policy.

Documentation

Files (dot)
  • SKILL.md 3.4 KB
    ---
    name: kaggle
    description: "Operate Kaggle competitions, datasets, kernels, and submissions."
    license: MIT
    metadata:
      kind: connector
      author: Médéric HURIER (Fmind)
      source: github.com/fmind/dot/tree/main/skills/kaggle
      created: "2026-09-16"
      updated: "2026-11-10"
    ---
    
    # Kaggle CLI
    
    Operate Kaggle competitions, datasets, kernels, and models via `kaggle` (auth, download bounds, and authority gates on submissions, remote runs, and publications).
    
    ## Workflow
    
    1. **Verify account**: `kaggle auth login` (OAuth), env `KAGGLE_API_TOKEN`, or legacy `~/.kaggle/kaggle.json`; never run `kaggle auth print-access-token`.
    1. **Inspect before downloading**: check bounded listings, `kaggle competitions submission-limits <slug> --json`, and `kaggle competitions pages list list <slug> --content --page-name rules` (double `list` required by the parser; reading rules never authorizes accepting them).
    1. **Download into gitignored `data/`** (accept competition rules in the browser first, else HTTP 403):
    
       ```bash
       kaggle competitions download <slug> -p data/ --unzip
       kaggle datasets download <owner>/<name> -p data/ --unzip
       ```
    
    1. **Check deadlines**: inspect `kaggle competitions pages list list <slug>` (`--page-name timeline`, else `description`) and offer—or schedule when asked—[gws](../gws/SKILL.md) Calendar reminders one week before entry/merger deadlines and on final submission day.
    1. **Audit kernels before `push`**: `kaggle kernels init -p <dir>` creates `kernel-metadata.json`. Before authorized `kaggle kernels push -p <dir>`, verify upload files, target ID, data sources, `is_private: true` (default private), internet, `--accelerator`, `-t` timeout, `--no-run` (push executes immediately otherwise), and `kaggle quota`; check `kernels status <owner>/<slug>` and `kernels output <owner>/<slug> -p out/`.
    1. **Authorize submissions** (each counts against daily caps and posts to the leaderboard; `--wait` exits non-zero on scoring error/timeout after consuming the slot, or check `competitions submission <ref>`):
    
       ```bash
       kaggle competitions submit <slug> -f submission.csv -m "<message>" --wait 900
       kaggle competitions submissions <slug> --format json --page-size 5
       ```
    
    1. **Authorize dataset publications**: `kaggle datasets create -p <dir>` defaults to private unless `--public` is approved; `datasets version -p <dir> -m "<message>"` keeps existing visibility (verify first, confirm license/audience, and pass history-destroying `-d/--delete-old-versions` only when requested).
    
    ## Gotchas
    
    - **Prefer `uv run kaggle`** when a project pins `kaggle` so the locked version runs over the global shim.
    - **Strip import-time `Warning:` lines** before JSON parsing (`-W` takes effect after import-time auth checks).
    
    ## Official Skills
    
    Upstream: `Kaggle/kaggle-cli` (CLI commands) and `Kaggle/kaggle-skills` (judging, exams, Kaggle Benchmarks), reviewed per [vendor-skill policy](../agent-project/references/vendor-skills.md).
    
    ## Documentation
    
    - [Kaggle CLI](https://github.com/Kaggle/kaggle-cli) · [Kaggle API](https://www.kaggle.com/docs/api)
    - Releases: [Kaggle CLI](https://github.com/Kaggle/kaggle-cli/releases)
    - Companion skills: [gws](../gws/SKILL.md) (calendar reminders), [python-stack](../python-stack/references/foundation/GUIDE.md) (project layout), [duckdb](../duckdb/SKILL.md) (inspect downloads), [hf](../hf/SKILL.md) (Hub models and datasets), [colab](../colab/SKILL.md) (rented accelerators).
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.