Data Artisan

AI-powered skills for building production-grade data pipelines, schemas, and systems—with Claude Code, Cursor, or any agent that supports Agent Skills.

The Problem

When you ask Claude Code to generate a database schema or data pipeline, it creates working code, but often misses the patterns that make systems production-ready:

• Missing indexes on high-cardinality columns
• No partition strategies for scale
• Incomplete data quality checks
• No lineage or observability patterns
• Designs that work at 1GB but break at 1TB

Data Engineering Skills teaches AI agents how to think like data engineers: designing for scale, reliability, and operations from day one.

What You Get

Each skill is a portable SKILL.md file that works with:

• Claude Code in Claude.ai
• Cursor (Claude Mode)
• Codex and other Agent Skills–compatible tools
• Direct paste into any coding agent

Available Skills

Skill	Install Name	What It Does
Schema Design Fundamentals	data-schema-design	Generate production-grade SQL schemas with indexing, partitioning, and data type strategies for Postgres/Snowflake/BigQuery
ETL Pattern Library	data-etl-patterns	Common ETL patterns: SCD Type 2, slowly changing facts, incremental loads, full refreshes with idempotency
Data Quality Framework	data-quality-checks	Generate dbt tests, Great Expectations, or SQL validation rules for completeness, uniqueness, referential integrity
Local AI Data Stack	local-ai-data	Optimize Claude Skills for local LLMs with proper context windows, fallback strategies, and model selection
DuckDB & Analytics	duckdb-analytics	Modern analytics patterns: columnar design, query optimization, analytical SQL best practices

Quick Start

Install with CLI

npx skills add https://github.com/Yuuzulight/data-engineering-skills --skill "data-schema-design"

Or Copy-Paste

1. Open skills/data-schema-design/SKILL.md
2. Copy the full file
3. Paste into Claude Code / Cursor / your agent of choice

Examples

Before: Naive Schema

CREATE TABLE users (
  id SERIAL PRIMARY KEY,
  name VARCHAR(255),
  email VARCHAR(255),
  created_at TIMESTAMP,
  updated_at TIMESTAMP
);

After: Production Schema (with skill)

CREATE TABLE users (
  id BIGSERIAL PRIMARY KEY,
  email VARCHAR(255) NOT NULL UNIQUE,
  name VARCHAR(255) NOT NULL,
  -- Denormalized domain for efficient filtering
  email_domain VARCHAR(255) NOT NULL,
  created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
  updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
);

-- High-cardinality search
CREATE INDEX idx_users_email ON users(email);
-- Domain-based queries (e.g., "all @company.com users")
CREATE INDEX idx_users_email_domain ON users(email_domain);
-- Time-range queries
CREATE INDEX idx_users_created_at ON users(created_at DESC);

-- Constraints for data quality
ALTER TABLE users ADD CONSTRAINT chk_email_format 
  CHECK (email ~* '^[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}$');

Repository Structure

data-engineering-skills/
├── README.md                           # This file
├── LICENSE                             # MIT
├── CHANGELOG.md                        # Version history
├── skills/
│   ├── data-schema-design/
│   │   └── SKILL.md                    # Main skill file
│   ├── data-etl-patterns/
│   │   └── SKILL.md
│   ├── data-quality-checks/
│   │   └── SKILL.md
│   ├── local-ai-data/
│   │   └── SKILL.md
│   └── duckdb-analytics/
│       └── SKILL.md
├── examples/
│   ├── schema-design-examples.md       # Real-world schema examples
│   ├── etl-patterns-examples.md        # ETL pattern implementations
│   └── data-quality-examples.md        # Quality check examples
├── research/
│   └── data-engineering-principles.md  # Background reading
└── scripts/
    └── skill.sh                        # CLI helper script

Why These Skills

Real Problems, Real Solutions

• Schema Design: Most AI-generated schemas are under-indexed. We fix that.
• ETL Patterns: Incremental loads and slowly-changing dimensions are complex. We teach the patterns.
• Data Quality: 80% of data issues come from quality problems upstream. We make them systemic.
• Local AI Data: As AI moves local, data stacks need to optimize for resource constraints. We show how.
• Analytics: Modern analytics (DuckDB, Trino, etc.) require different thinking than OLTP. We bridge the gap.

Who Should Use This

• Data Engineers who want Claude Code to generate better database designs
• ML Engineers who need production data pipelines but aren't data specialists
• Startups building data infrastructure quickly without hiring a full data team
• Data Teams who want to standardize how AI generates data systems

Getting Help

• Have a question? Open a GitHub Discussion
• Found a bug or gap? Open an Issue
• Want to contribute? See CONTRIBUTING.md

Support Development

Building and maintaining these skills takes time. If they're saving you hours on database design, ETL pipelines, or data quality, consider sponsoring development.

Support Data Artisan on GitHub Sponsors

Sponsors get:

• Early access to new skills
• Priority for feature requests
• Recognition in README
• Direct feedback channel

Sponsors

Roadmap

• Kafka & Streaming Skills – streaming data patterns, exactly-once semantics, backpressure handling
• dbt Advanced Patterns – macro patterns, custom tests, performance optimization
• Data Governance – PII masking, lineage tracking, access control patterns
• Cloud Cost Optimization – Snowflake cost patterns, BigQuery slot management, data lifecycle policies
• Real-time Analytics – event streaming, dimensional modeling for real-time

License

MIT License © 2026. See LICENSE for details.

Disclaimer

These skills are designed to enhance AI-generated code, not replace human data engineers. Always review generated code, test thoroughly, and follow your organization's data governance policies.

---

Questions? Open a discussion or tweet @Yuuzulight