8000
Skip to content

yonikremer/claude-data-skills

 
 

Repository files navigation

Claude Data Skills 🐍📊

A professional-grade collection of data science, analysis, and engineering skills and scripts for AI-assisted development.

PyPI Version License: MIT Python 3.9+

Overview

claude-data-skills is a comprehensive library designed to enhance AI-assisted data workflows. It provides a structured collection of "skills"—reusable, idiomatic patterns and scripts for everything from advanced standard library usage to complex machine learning pipelines and professional development workflows.

Key Features

  • 🚀 Professional Python Core: Unified expert guide for PEP-8, Pydantic, Pytest, and high-performance parallelism.
  • 📊 Data Analysis Pro: Consolidated power-user guide for NumPy, Pandas, and Polars. Unified strategy for scaling from KB to 100GB+.
  • 🕸️ Full Spectrum Graph Sieve (GraphRAG): Advanced agentic workflow for extracting relationship-aware domain knowledge from internal docs (.docx, .one, .msg, .pdf). Now powered by the external graph-sieve package.
  • ⚡ Superpowers Workflow: Integrated skills for brainstorming, TDD, systematic debugging, and plan execution.
  • 🛡️ Data Safety First: Built-in guardrails to prevent accidental data loss or corruption during autonomous execution.
  • 📈 Visualization Pro: Expert guide for Plotly (interactive), Dash (dashboards), and Seaborn (static stats).
  • 🗄️ Database Pro: Unified access for SQL (Postgres), SQLAlchemy (ORM), Elasticsearch, and S3.
  • 📁 Document Processing Pro: Consolidated expert guide for PDF, Word (DOCX), Excel (XLSX), and PowerPoint (PPTX).
  • 🔬 Scientific Research Suite: Unified guide for the entire scientific lifecycle: brainstorming, writing (IMRAD), and peer review.
  • 🔄 Legacy Migration Suite: Specialized patterns for migrating C#, MATLAB, and Python 2 code to modern Python ( 3.9+).

Installation

Install the package directly from PyPI:

pip install claude-data-skills

Quick Start

Post-Installation Setup

After installing the package, run the following command to copy the necessary skills files to your user's Claude home directory (~/.claude/skills):

setup-claude-skills

Using the CLI

The package includes several built-in commands. For example, to run the standard library demonstration:

stdlib-demo

Importing Skills

You can import advanced utility patterns directly into your own scripts:

from skills.python_dev.python_stdlib_pro.scripts.stdlib_demo import test_pathlib

# Run a verified pathlib pattern
test_pathlib()

Core Principles

  • Resource Aware: Every intensive task starts with hardware resource validation.
  • LLM Optimized: Scripts are dense, idiomatic, and contain strict guardrails for local/open-source LLMs.
  • Atomic Operations: Prevents file corruption by using temp-and-replace patterns for all writes.

Project Structure

skills/
├── core-workflow/          # Brainstorming, TDD, Debugging, Plans
├── data-analysis/          # Data Analysis Pro (NumPy, Pandas, Polars), Geopandas
├── data-sources/           # Database Pro (Postgres, SQLAlchemy, ES, S3)
├── machine-learning/       # ML-Classical, ML-Deep-Learning, PyMC
├── python-dev/             # Python Core Pro, Legacy Migration Suite, Logic Recovery
├── scientific-workflow/    # Scientific Research Suite
└── unstructured-data/      # Document Processing Pro, Binary Data Parsing

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

Created and maintained by Yoni Kremer.

About

A set of ready to use Agent Skills for research, science, engineering, analysis, finance and writing.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages

  • Python 70.4%
  • HTML 20.4%
  • TeX 8.7%
  • Other 0.5%
0