What does this skill do, and when should you use it?
This is an independent, unofficial Agent Skill repository that gives SKILL-compatible coding agents full Kaggle coverage. It handles credential setup, competition reports and evaluation summaries, dataset/model download and publishing, notebook push and remote execution with polling, forum and writeup retrieval, benchmark workflows, and badge collection. The repo is organized as workflow modules with concrete scripts, wraps all Kaggle-supplied text in untrusted-content markers, and ships a security test suite. It is distributed via skills.sh, ClawHub, the Codex plugin marketplace, and the Claude Code plugin marketplace, and is compatible with 35+ agents.
- Reads and summarizes competition pages: rules, evaluation metric, leaderboards, and landscape reports
- Downloads and publishes datasets and models via kagglehub or the kaggle CLI
- Publishes notebooks to Kaggle Kernels, executes them remotely, polls status, and downloads outputs
- Retrieves forum topics, leaderboard solution writeups, and hackathon writeups (preserving role-denial responses as evidence)
- Runs Kaggle benchmarks CLI workflows for task creation, model runs, and status
- Executes badge collection phases and streak helpers, always with a dry-run first
- A data-science competitor asks the agent to summarize rules and the evaluation metric for the Titanic competition before submitting
- A researcher pulls the top-K leaderboard writeups for a competition and groups hackathon writeups by track
- A participant downloads a competition dataset and prepares it for a local notebook
- A user pushes a local notebook to Kaggle Kernels, waits for execution, and retrieves the output
- A Kaggle user checks which badges are still earnable via API activity, previewing with a dry run first
- Generic ML, dataset, or GPU/TPU work unrelated to Kaggle — the skill description explicitly instructs not to activate
- Users without a Kaggle account or KAGGLE_API_TOKEN — the credential check gates every operation
- Offline environments — outbound HTTPS to api.kaggle.com, www.kaggle.com, and storage.googleapis.com is required
How do you install this skill?
- Badge collection and competition submission flows directly modify the user's Kaggle account (creating private resources, consuming submission slots) — confirm each action before execution.
- Submitted prediction files are placeholders (e.g. all-zero Titanic predictions) and could affect account standing or violate some competition rules; do not use against real competitions casually.
- Core functionality depends on api.kaggle.com and www.kaggle.com, which may be unreachable from mainland-China networks without additional networking.
- This is a static review with no execution; minor inconsistencies exist (badge count 55 vs 59) — verify details before use.
- Publisher identity is unverified by the FollowSkills registry (unknown, not suspicious); additional due diligence is advised for enterprise adoption.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
Python 3.11+kagglehub>=1.0.0kaggle>=2.2.3kagglesdk<1.0requestspython-dotenvKAGGLE_API_TOKEN
Pick one install route for your host:
Claude Code
/plugin marketplace add shepsci/kaggle-skill
/plugin install kaggle@shepsciCodex
codex plugin marketplace add shepsci/kaggle-skill --ref main
codex plugin add kaggle@shepsciskills.sh
npx skills add shepsci/kaggle-skillClawHub
clawhub install kaggleManual clone
git clone https://github.com/shepsci/kaggle-skill.git
cd kaggle-skill
python3 -m pip install "kagglehub>=1.0.0" "kaggle>=2.2.3" "kagglesdk>=0.1.33,<1.0" python-dotenv requestsHow do you use this skill?
Once installed, send your agent any of these to trigger it:
- Use the Kaggle skill to summarize the rules and evaluation metric for the Titanic competition.
- Retrieve and preview the writeups from the top 3 ranked submissions in the Vesuvius Challenge surface detection competition.
- Download the Kaggle dataset owner/dataset-name into ./data and prepare it for a notebook.
- Push this notebook to Kaggle Kernels and tell me when it finishes, then download the output.
The skill triggers when you explicitly mention Kaggle — a URL, competition, dataset, writeup, badge, or account setup; it deliberately stays dormant for generic ML tasks. Run the credential checker first, then hand the request to your agent, which routes to the matching module (setup/competitions/datasets/models/notebooks/discussions/benchmarks/badges). Write actions (submissions, publishing, badge phases) default to read-only/dry-run and require explicit user intent before executing. Example credential setup:
mkdir -p ~/.kaggle
printf '%s\n' 'YOUR_TOKEN' > ~/.kaggle/access_token
chmod 600 ~/.kaggle/access_token
python3 modules/setup/scripts/check_all_credentials.pyWhat are this skill's strengths and limitations?
- Full Kaggle lifecycle coverage: credentials, competitions, datasets, models, notebooks, discussions, benchmarks, badges
- Solid security design: no dynamic eval, no credential leakage, untrusted-content wrapping, zip-slip protection — each backed by tests
- Write actions force dry-run-first plus explicit user intent, preventing accidental submissions or publishes
- Broad distribution: documented install commands for Claude Code, Codex, skills.sh, and ClawHub
- Independent and unofficial — not affiliated with Kaggle or Google, so platform changes can break scripts
- The competitions module's SPA-scraping steps assume the host agent provides Playwright MCP tools; the skill does not bundle them
- Requires Python 3.11+ and several pip dependencies — heavier setup than a prompt-only skill
- Automated badge collection may run against Kaggle's stance on automation; users should assess that risk themselves
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Kaggle Skill this page | 61 · Recommended | ★ 103 | 10d ago | MIT |
| Bright Data Web MCP Skill | 51 · Use with care | ★ 38k | 13d ago | MIT |
| Engram Persistent Memory Protocol | 68 · Recommended | ★ 7.2k | 11d ago | MIT |
| Node Modules Inspector — Dependency Audit Skill | 67 · Recommended | ★ 3k | 1mo ago | MIT |
| e2e Verify — End-to-End Change Verification Skill | 64 · Recommended | ★ 10k | 11d ago | Apache-2.0 |
Compared with using the kaggle CLI or kagglehub directly, this skill wraps the commands into agent-dispatchable workflows and adds credential checks, a dry-run safety policy, and untrusted-content guards. Compared with raw Kaggle MCP server tools, it adds higher-level orchestration such as writeup discovery and badge workflows, and its bundled .mcp. configures Kaggle's remote MCP endpoint (live inventory verified at 70 tools).
How did FollowSkills review this skill?
Evidence shows solid security design: scoped allowed-tools, explicit-intent gating for writes, dry-run defaults, credential non-echoing, zip-slip protection, untrusted-content markers, with test filenames cited (full test bodies not read). Deductions: submissions and badge farming directly modify the user's Kaggle account and submit placeholder predictions (e.g. all-zero Titanic), carrying quota/visibility risk despite confirmation requirements; publisher identity unverified.
Scripts are internally consistent with reasonable error handling, status tracking (badge_tracker), readable failure messages, and skip/fallback logic. Deductions: static review only — key paths (CLI output parsing, kagglehub upload, Playwright phase) depend on external services with brittle format assumptions; badge-catalog has internal inconsistency (55 vs 59); one script is truncated in evidence, so completeness cannot be confirmed.
Precise trigger description (explicitly excludes non-Kaggle tasks), clear module routing, broad compatibility claims, and a clear non-fit boundary (17 non-automatable badges). Deductions: core function depends entirely on overseas services (api.kaggle.com, www.kaggle.com), a real mainland-China reachability limitation not discussed; Playwright MCP tools assumed from the host agent add environmental conditions.
Well-layered docs (SKILL.md → module READMEs → references), MIT license, version 2.4.0 consistent with pyproject, explicit unofficial disclaimer, SECURITY.md and privacy statement. Deductions: no CHANGELOG or explicit maintenance commitment; minor factual inconsistencies (badge counts); troubleshooting/testing coverage partly asserted only via README.
Core workflows (competition page summaries, dataset downloads, discussion retrieval) have concrete command examples and demo screencasts claimed; marginal value over manual browsing is plausible. Deductions: static review cannot verify output quality; badge automation has limited practical benefit with some steps marked manual; demos are author-recorded and not independently verifiable.
Repo includes a test directory, a security-test matrix, pytest markers (live/destructive/manual), and reproducible command paths. Deductions: actual test bodies and CI config were not present in the evidence; demo casts and the '70 MCP tools verified 2026-07-03' claim are author assertions, so more than 5 cannot be awarded on a static read.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →