Load and Profile a CSV

In this tutorial you'll watch the MAGIC Data Agent load a CSV file and run a full data-quality profile on it. The agent automatically detects the file format, loads it into memory, and then routes to the profiling skill to score quality and surface critical issues — all from a single natural-language prompt.

This is the recommended starting point if you're new to MAGIC Data Agent Skills. It uses two skills (magic-data-loading and magic-data-profiling) and takes under two minutes end-to-end.

Setup

npx @votee-ai/magic-data-agent-skills init

mkdir magic-tutorial && cd magic-tutorial

curl -O https://docs.votee.ai/tutorials/data/sales_data.csv

Sample Data

This tutorial uses a 515-row sales dataset with intentional quality issues — mixed date formats, duplicate rows, missing values, and extreme outliers. Download it to follow along:

The Agent Trace

The interactive viewer below shows every step the agent took — from receiving your prompt to delivering the quality report. Click any step to expand its details, or use ← → arrow keys to navigate.

{
  "file_path": "data/sales_data.csv"
}

{
  "rows": 515,
  "columns": 12,
  "encoding": "UTF-8",
  "format": "CSV"
}

wc -l data/sales_data.csv && file data/sales_data.csv

{
  "dataframe_shape": [
    515,
    12
  ]
}

{
  "quality_score": 86.9,
  "grade": "B",
  "duplicates": 15,
  "date_formats": 3,
  "quantity_outliers": 5,
  "missing_email_pct": "13.0%",
  "missing_phone_pct": "8.2%",
  "missing_notes_pct": "89.7%",
  "revenue_consistency": "100%",
  "email_shared_across_reps": true
}

import pandas as pd
import numpy as np
import re

df = pd.read_csv('data/sales_data.csv')
# Quality dimensions: completeness, uniqueness,
# consistency (date formats, email mapping),
# validity (outliers, revenue check)

{
  "columns": [
    "date",
    "quantity"
  ]
}

{
  "date_formats": {
    "YYYY-MM-DD": 410,
    "MM/DD/YYYY": 69,
    "DD-Mon-YYYY": 36
  },
  "quantity_outliers": [
    {
      "order_id": "ORD-1389",
      "quantity": 15862,
      "revenue": "$1.6M"
    },
    {
      "order_id": "ORD-1444",
      "quantity": 12353,
      "revenue": "$1.7M"
    },
    {
      "order_id": "ORD-1256",
      "quantity": 11581
    },
    {
      "order_id": "ORD-1042",
      "quantity": 9262
    },
    {
      "order_id": "ORD-1137",
      "quantity": 5741
    }
  ]
}

# Date format detection
mask_iso = df['date'].str.match(r'^\d{4}-\d{2}-\d{2}$')
mask_us = df['date'].str.match(r'^\d{2}/\d{2}/\d{4}$')
mask_dmy = df['date'].str.match(r'^\d{2}-[A-Z][a-z]{2}-\d{4}$')

Data Flow

What Happened

Key Issues Found

Try It Yourself

Copy this prompt and paste it into your MAGIC session:

Load data/sales_data.csv and tell me about the data quality issues

The agent will:

1. Detect the file format and encoding
2. Load the data into memory
3. Transition automatically to the profiling skill
4. Return a quality score with a prioritized list of issues

What You Learned

• The agent automatically selects skills based on keywords in your prompt — you don't need to name skills explicitly.
• Skill transitions happen seamlessly: magic-data-loading hands off to magic-data-profiling without any extra instruction from you.
• The agent runs multiple analysis passes (dates, nulls, duplicates, outliers, cross-column validation) in a single profiling session.
• Missing values, duplicates, and outliers are all surfaced in a single pass with severity ratings to help you prioritize.

Alternate Approaches

Next Steps