● 社群 shepsci ⚡ 即開即用

Kaggle Skill

Name: Kaggle Skill (Claude Skill)
Author: shepsci

Complete Kaggle integration — competition reports, dataset/model downloads, notebook execution.

A skill that wires Kaggle into Claude Code: list competitions, download datasets/models, submit predictions, fetch notebook executions, track badges. Works across Claude Code, Cursor, Codex, Gemini CLI through skills.sh.

▶ 觀看演示 📦 安裝 💡 使用場景

為什麼要用

核心特性

Competition browsing + submission
Dataset and model download with versioning
Notebook execution status + outputs
Badge progress tracking
Multi-agent compatibility via skills.sh standard

即時演示

實際使用效果

kaggle-skill.replay ▶ 就緒

0/0

安裝

選擇你的客戶端

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "kaggle-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/shepsci/kaggle-skill",
        "~/.claude/skills/kaggle-skill"
      ],
      "_inferred": true
    }
  }
}

開啟 Claude Desktop → Settings → Developer → Edit Config。儲存後重啟應用。

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "kaggle-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/shepsci/kaggle-skill",
        "~/.claude/skills/kaggle-skill"
      ],
      "_inferred": true
    }
  }
}

Cursor 使用與 Claude Desktop 相同的 mcpServers 格式。專案級設定優先於全域。

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "kaggle-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/shepsci/kaggle-skill",
        "~/.claude/skills/kaggle-skill"
      ],
      "_inferred": true
    }
  }
}

點擊 Cline 側欄中的 MCP Servers 圖示，然後選 "Edit Configuration"。

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "kaggle-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/shepsci/kaggle-skill",
        "~/.claude/skills/kaggle-skill"
      ],
      "_inferred": true
    }
  }
}

格式與 Claude Desktop 相同。重啟 Windsurf 生效。

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "kaggle-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/shepsci/kaggle-skill",
        "~/.claude/skills/kaggle-skill"
      ]
    }
  ]
}

Continue 使用伺服器物件陣列，而非映射。

~/.config/zed/settings.json

{
  "context_servers": {
    "kaggle-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/shepsci/kaggle-skill",
          "~/.claude/skills/kaggle-skill"
        ]
      }
    }
  }
}

加入 context_servers。Zed 儲存後熱重載。

claude mcp add kaggle-skill -- git clone https://github.com/shepsci/kaggle-skill ~/.claude/skills/kaggle-skill

一行命令搞定。用 claude mcp list 驗證，claude mcp remove 移除。

使用場景

實戰用法： Kaggle Skill

Pull a competition dataset, train a baseline, submit predictions

👤 Kagglers iterating fast ⏱ ~15 min intermediate

何時使用： You want to get a first submission on the leaderboard quickly.

前置條件

Server/skill installed and authenticated — See repo README

步驟

Pull data

Use the Kaggle skill to download the latest version of competition 'titanic' to ./data.✓ 已複製

→ Dataset downloaded; files listed
Submit

After my notebook runs, submit submission.csv with message 'baseline LR'.✓ 已複製

→ Submission id + public score when scored

結果： A real submission on the leaderboard with one chat.

注意事項

Daily submission limits per competition — usually 5–10. The skill warns when you're close. — Daily submission limits per competition — usually 5–10. The skill warns when you're close.

搭配使用： filesystem

組合

與其他 MCP 搭配，撬動十倍槓桿

kaggle-skill + filesystem

Pair with filesystem

Use this skill together with filesystem.✓ 已複製

工具

此 MCP 暴露的能力

工具	輸入參數	何時呼叫	成本
list_competitions	category, status	Find active competitions	1 API call
download_dataset	ref, version	Pull a dataset locally	Disk I/O
submit	competition, file, message	Submit predictions	1 API call

成本與限制

運行它的成本

API 配額: Kaggle rate limits apply
每次呼叫 Token 數: Varies
費用: Free for public competitions
提示: Cache downloads.

安全

權限、密鑰、影響範圍

憑證儲存： ~/.kaggle/kaggle.json with chmod 600

資料出站： kaggle.com

故障排查

常見錯誤與修復

403 Forbidden on download

Accept the competition's rules on Kaggle first via the web UI. The CLI cannot accept rules.

驗證： Visit competition page

kaggle.json missing

Place your API token at ~/.kaggle/kaggle.json with chmod 600.

驗證： kaggle competitions list

替代方案

Kaggle Skill 對比其他方案

替代方案	何時用它替代	權衡
Kaggle CLI directly	You want raw CLI	No agent integration

資源

📖 閱讀 GitHub 上的官方 README

🐙 查看未解決的 issue

🔍 瀏覽全部 400+ MCP 伺服器和 Skills