wenmin-wu/ds-skills
Developer toolsこのページは AI により自動翻訳されています。技術的な詳細は原文をご確認ください。 原文を見る
Data Science Skills Distilled from Awesome Kaggle Notebooks.
概要
Data Science Skills Distilled from Awesome Kaggle Notebooks. 535 reusable techniques across 5 domains, each extracted from top-voted Kaggle competition solutions. Every skill is a self-contained SKILL.md — ready to be loaded by Claude Code, Cursor, Codex, or any AI coding agent. More skills coming soon — 600+ competitions in the pipeline, with new skills distilled regularly via an automated agent. Star this repo to stay updated. Full list: run ds-skills list or browse ds-skills.com. See Changelog for what's new. Copy this repo URL into your AI coding agent (Claude Code, Cursor, Codex, etc.) and ask it to install ds-skills: Your agent can read the README, install the CLI, and set everything up by itself. Add --json to any command for structured JSON output (agent-friendly). If you prefer to clone the repo and copy skills locally:
README
以下は原文 README の抜粋です。原文を見る。
ds-skills
Data Science Skills Distilled from Awesome Kaggle Notebooks.
535 reusable techniques across 5 domains, each extracted from top-voted Kaggle competition solutions. Every skill is a self-contained SKILL.md — ready to be loaded by Claude Code, Cursor, Codex, or any AI coding agent.
More skills coming soon — 600+ competitions in the pipeline, with new skills distilled regularly via an automated agent. Star this repo to stay updated.
Skills
| Domain | Count | Examples |
|---|---|---|
| tabular | 123 | hierarchical-rule-engine, opponent-avoidance-scoring, ball-landing-prediction, chained-target-prediction |
| nlp | 119 | lda-topic-modeling, taxonomy-context-enrichment, learned-attention-pooling, domain-special-token-embedding |
| cv | 182 | stroke-temporal-color-rendering, conv1d-lstm-stroke-classifier, deterministic-hash-partitioning, chunked-csv-image-generator |
| timeseries | 61 | multi-lag-target-features, periodogram-seasonality-detection, kaggle-api-streaming-inference, ratio-target-for-smape |
| llm | 50 | perplexity-prompt-ranking, low-meaning-input-substitution, 4bit-nf4-double-quantization, inverse-task-prompt-template |
Full list: run ds-skills list or browse ds-skills.com. See Changelog for what’s new.
Install skills to your agent
Option 1: Let your AI agent do it
Copy this repo URL into your AI coding agent (Claude Code, Cursor, Codex, etc.) and ask it to install ds-skills:
https://github.com/wenmin-wu/ds-skills
Your agent can read the README, install the CLI, and set everything up by itself.
Option 2: ds-skills CLI
pip install ds-skills-cli
ds-skills install --agent claude-code # or: cursor, codex
ds-skills setup --agent claude-code # teach your agent the CLI
More commands:
ds-skills install nlp/deberta-classification --agent claude-code # single skill
ds-skills install --agent claude-code --domain tabular # single domain
ds-skills list # list all skills
ds-skills search "gradient boosting" # search by keyword
ds-skills show nlp/deberta-classification # preview a skill
ds-skills stats # aggregate stats
Add --json to any command for structured JSON output (agent-friendly).
Option 2: Local copy script
If you prefer to clone the repo and copy skills locally:
git clone https://github.com/wenmin-wu/ds-skills.git
cd ds-skills
# Copy to your agent's skill directory
scripts/skills-copy --dest ~/.claude/skills # Claude Code
scripts/skills-copy --dest ~/.cursor/rules # Cursor
scripts/skills-copy --dest ~/.codex/skills # Codex
Skills are flattened on copy: skills/nlp/deberta-classification/ becomes nlp-deberta-classification/.
Structure
skills///SKILL.md
推奨ツール
別のキーワードを試すか、フィルタを外してください。
インストール
npx skillfish add wenmin-wu/ds-skills