AI Data Security Demo — Security skill for Claude Code

AI Data Security for Data Professionals

Companion repository for the [KC Labs AI](https://www.youtube.com/@kylechalmersdataai) video series on AI data security for data professionals.

A practical framework and hands-on toolkit for understanding what happens to your data when you use AI coding tools like Claude Code, Cursor, and GitHub Copilot — especially when you connect them to your databases.

**Now available as a Claude Code plugin:** [**AI Data Security**](https://github.com/kyle-chalmers/ai-data-security) turns the audit side of this series into runnable skills. It scans a repo, your AI-tool configuration, and a Postgres/Snowflake warehouse against the framework below, and reports cited, severity-ranked findings — read-only, every finding mapped to a standard.

claude plugin marketplace add kyle-chalmers/ai-data-security
claude plugin install ai-data-security@ai-data-security
# then: /ai-data-security:security-audit 

This repo is organized as chapters, one per topic. Each chapter pairs with a video and can be used independently.

Chapter	Video	What it covers
Top-level (this README)	What Your AI Coding Tools Actually Do with Your Data	4-tier sensitivity framework, plan tier comparison, gitleaks secret scanning
`snowflake/`	How to Safely Connect AI to Your Data	Starting state for the Snowflake build — scoped role + masking view + per-row salting walkthrough

Key setup guides

• Snowflake service-user RSA key-pair setup — generate, register, and wire up the dedicated AI service user (kchalmers_ai) the Snowflake chapter authenticates as.


---

The Problem

Data professionals use AI coding tools daily with real company data, often without understanding what happens to that data. Key risks:

• Plan tier determines exposure: Free/consumer plans m