MAGIC Agent Skills · Open source

Turn any Agent into a data expert

Open-source skill suites that drop straight into the tools you already use. Install once, deploy everywhere — and give Claude Code, Cursor, Windsurf and 27+ others real data-science and linguistic expertise.

Apache 2.030+ AI assistants13 slash commands

Works with Claude Code

+ 25 more

magic-data-agent-skills

MAGIC Agent Skills CLI installing data science skills

Architecture

The 3-layer skill architecture

Every skill is built from three layers the agent reads — then writes its own code from.

Layer 01

Domain Knowledge

What the agent must understand about your task — when to activate, plus the rules, constraints, and anti-patterns to avoid.

Layer 02

Code Patterns

Seed templates and reference implementations the agent studies — then writes its own custom code for your data, never run as-is.

Layer 03

Procedures

Step-by-step workflows. Reference scripts in scripts/ show the approach; the agent adapts them into code for the task at hand.

Installation

Quickstart

One command, zero config. The installer auto-detects your tools and drops in all 12 skills and 13 slash commands — no API keys, no boilerplate, no lock-in.

Just need an AI coding assistant

Claude Code, Cursor, Windsurf, or 27+ others — no other setup required.

install

# Interactive install — auto-detects your tools
$ npx @votee-ai/magic-data-agent-skills init
 
# Or target specific tools
$ npx @votee-ai/magic-data-agent-skills init --tools claude,cursor,windsurf

Deploys 12 skills and 13 slash commands, auto-configured for each tool.

Skill suites

Two suites. One open ecosystem.

Pick a domain and your assistant becomes an expert in it — with more suites on the way.

Data Science · 12 skills

Data Agent Skills

Production-grade skills covering the complete data lifecycle — ingestion, analysis, cleaning, transformation, synthetic generation, testing, visualization, and reporting.

Multi-format loading with automatic encoding detection
Quality scoring, profiling, and outlier detection
Schema validation and constraint verification
LLM-powered synthetic data generation
Statistical testing and correlation analysis
Automated visualization and Markdown reports

Low-Resource NLP · 18 skills

Linguistic Agent Skills

Specialized skills for multilingual natural language processing — built for low-resource languages, corpus preparation, and deep linguistic examination.

Low-resource and multilingual language support
Corpus preparation and curation
Linguistic analysis and examination tasks

How it works

The full data lifecycle, automated

Your agent calls MAGIC skills, which ingest your data and run one orchestrated pipeline — from raw files to a finished, cited report.

① AI Agents · your assistant

Claude Code Cursor

Cursor

Windsurf

GitHub Copilot Codex

Codex

Antigravity+ 24 more

Call MAGIC skills

② Accessing MAGIC Skills

MAGIC Agent Skills

12 data skills · 13 slash commands

Runs on

Inject data

③ Data Injection

CSVJSON

Parquet

ExcelDatabases HuggingFace Hub

HuggingFace Hub

Pipeline orchestration

Ingest

CSV · Parquet · JSON · DB · HF Hub

Profile

Quality scoring & distributions

Clean

Missing values & dedupe

Validate

Schema & constraints

Discover

Patterns & relationships

Transform

Pivot, reshape & merge

Synthesize

LLM-powered data gen

Test

Stats & correlation

Visualize

Static & interactive

Report

Structured Markdown

Compatibility

Plays nicely with your stack

One configuration, no redundant setup — the same skills, ready in every assistant you reach for.

Claude Code Cursor

Cursor

Windsurf

GitHub Copilot Codex

Codex

Antigravity OpenAI

OpenAI+ 23 more