SkillFoxx

ds-ai-coding-skills

A template for data scientists: AGENTS.md, CLAUDE.md and skills for Python, SQL, Polars and model diagnostics.

  • 115 ★ on the repository
  • Last commit
  • MIT
Skill

Minimal access

Why this level

  • The skills are instructions for data and code work, and the secret and raw-data checks run locally.
All reasons and checks

atsushi-green/ds-ai-coding-skillsAuthor:atsushi-green

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/ds-ai-coding-skills

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Minimal accessThe tool runs nothing and makes no network calls.

  • The skills are instructions for data and code work, and the secret and raw-data checks run locally.

This is third-party code. Review the repository files before installing.

What it does

ds-ai-coding-skills gathers agent working rules for a data science project in one place. A single text lives in CLAUDE.md and AGENTS.md: hard rules, shared commands and skill routing. The skills themselves live in `.claude/skills/`, and `.agents/skills` points to them by symlink, so Claude Code, GitHub Copilot and Codex read the same thing. Base skills cover Python and uv, SQL, Polars, paths and I/O, notebooks, safe data handling, reports and pull request preparation. A separate `*-diagnostics` series for statistics, causality, predictive models and simulation requires a set of plots and values with pass criteria every time, so the agent does not stop at R-squared.

Who it is for. Data scientists and ML engineers working on Python and Polars projects in Claude Code, Copilot or Codex who want unified rules and mandatory diagnostics.

Good fit when

  • A new analytics project starts from a ready structure for data, notebooks and reports.
  • An agent picked a model but did not show residuals and calibration, and diagnostics by method must be demanded.
  • A repository must be protected from committing raw data and secrets by the checks in `scripts/`.
  • A team uses different AI tools and needs one source of rules.

Not a fit when

  • The skill text is Japanese and plans are written in Japanese; other languages require translation.
  • The project is not Python: the skills target uv, Polars and ruff.
  • You need one ready skill without a project template: the set works as a repository scaffold.

Example request

Pick a model for the customer churn table and show diagnostics: residuals, calibration and feature importance following the skill's method.

Limitations

Instructions and documentation are in Japanese, and some skills require writing the plan in Japanese. The repository is a project template: Python 3.11 and uv are needed, and commands like `uv run pytest` and `bash scripts/run_quality_checks.sh` belong to the template itself. There is no skill installer; copy the skills into your project together with the router or links between files will break. The diagnostic skills require plots, so visualization libraries must be present. Last commit 4 October 2026.

How to disable. Delete the copied folders from `.claude/skills/`, the `.agents/skills` link and the CLAUDE.md and AGENTS.md files if you took them from the template.

FAQ

How do I attach the skills to a project?

Copy `.claude/skills/` and the router CLAUDE.md or AGENTS.md into your repository. For Codex add the `.agents/skills` link pointing to `.claude/skills`.

What are the diagnostics skills?

A series of skills by method: statistical inference, causality, predictive models, simulation. Each requires a set of plots and values with an acceptance criterion.

Which quality checks exist?

Scripts look for raw data in a commit, secret patterns and drift between CLAUDE.md and AGENTS.md. Run `bash scripts/run_quality_checks.sh`.

Does the set work in GitHub Copilot?

Yes. Copilot reads AGENTS.md and takes detailed instructions from the skills. For JetBrains and Visual Studio add `.github/copilot-instructions.md` with the same text.

Is Japanese needed?

It helps for reading the rules. The skills are written in Japanese and the agent drafts the analysis plan in Japanese until you translate the files.

Discussion

No questions or reviews yet.

Skills

31 skills for Claude Code and Codex for the analyst workflow: data checks, cohorts, funnels, A/B tests, dashboards, summaries

SkillMinimal access529

17 skills for computational biology: RNA-seq, single-cell, protein structure, docking, literature and figures

SkillExtended access75

Around 200 skills for Claude Code on bioinformatics, chemistry and biostatistics, from RNA-seq analysis to molecular docking

SkillExtended access388

High-Stakes Analytics Lab

High-Stakes Analytics & Decision Lab

An evidence-gated analytics skill for high-stakes decisions: profiles data, picks a method the evidence supports, can answer with a data request

SkillExtended access1K

16 Claude Code skills for data engineers: source profiling, failure investigation, change review and documentation as web pages.

PluginExtended access98

README in short

The document moves from links to the author's articles and uv setup to the project directory layout and the single-source principle for instructions. Tables of skills and the diagnostics series follow.

README badge

Original SKILL.md

---
name: plan-analysis
description: 分析を始める前に構造化された分析計画を作成する
disable-model-invocation: true
---

# Skill: Plan Analysis

コードを書く前に、構造化された分析計画を作成する。

## 確認・整理する項目

1. **Objective**: どんな問いに答えようとしているか。
2. **Data sources**: 使用するデータ（テーブル、ファイル、API）。
3. **Unit of analysis**: 1行が何を表すか。
4. **Key metrics**: どの指標を計算するか、その定義。
5. **Risks**: 何が問題になりうるか（データ品質、リーケージ、バイアス、欠損値など）。
6. **Validation**: 結果をどう検証するか。
7. **Outputs**: 期待される成果物（テーブル、チャート、レポート、モデルなど）。

計画は日本語でまとめる。プロジェクト固有の文脈については `docs/agent/metrics-and-definitions.md` と `docs/agent/data-catalog.md` を参照する。

## 分析ワークフロー全体

この skill は下記の1に相当する。計画時には後続フェーズまで見通しておく。

1. **分析開始前** — 分析目的と意思決定への影響を確認する / 必要なデータの有無と品質を確認する / 分析計画を作成する（この skill）/ 指標定義を `docs/agent/metrics-and-definitions.md` で確認する
2. **データ探索** — 基本統計量・欠損値・異常値を確認する / 粒度とキーの一意性を確認する / 結果をNotebookに記録する（`run-eda` skill）
3. **分析・検証** — 分析ロジックを実装する / 既知の事実との整合性で妥当性を確認する / エッジケースを検討する（`run-modeling` skill、手法に応じた `*-diagnostics` skill）
4. **成果物作成** — 図表を `outputs/` に保存する / 分析結果をまとめる（`summarize-analysis` skill）/ 再現手順を記録する
5. **レビュー** — 品質チェックを実行する（`bash scripts/run_quality_checks.sh`）/ PR概要を作成する（`prepare-pr` skill）/ レビュアーに依頼する

Report a problem