Analyze structured data files (CSV, JSON) and produce clear, actionable reports.
When to Use
- User asks to "summarize", "analyze", or "describe" a data file
- User wants statistics: row counts, column types, means, medians, distributions
- User needs to find anomalies or outliers in numeric data
- User requests a formatted report or table from raw data
When NOT to Use
- Unstructured text files (use a text-processing skill instead)
- Database queries (use a database skill)
- Image or binary file analysis
- Real-time or streaming data
Workflow
- Identify the data file — look for CSV or JSON files in the workspace
- Run the analysis script for deterministic statistics:
python3 scripts/analyze_csv.py <file.csv> - Interpret the results — add context about what the numbers mean
- Format the output — use markdown tables for readability
Output Format
Always structure your response as:
- Overview — file name, row count, column count
- Column Summary — for each column: type, non-null count, unique values
- Statistics — for numeric columns: min, max, mean, median, std dev
- Anomalies — any outliers (values beyond 2 standard deviations)
- Recommendations — actionable insights based on the data
Important Notes
- Always report exact numbers, not approximations
- If a CSV has headers, use them as column names
- For large files (>10k rows), report a sample and note the total
- Never modify the source data file