| name | data-analysis |
| description | Analyze structured datasets (CSV, databases, numerical data). Triggers on 'analyze data', 'compute statistics', 'find patterns in data'. NOT for qualitative research — use research/analysis. |
| allowed-tools | Bash, Read, Write |
When to Use
- Computing descriptive statistics on structured data (CSV, TSV, databases)
- Identifying trends, correlations, or anomalies in numerical datasets
- Testing hypotheses with data evidence
- Segmenting or grouping records by attributes
When NOT to Use
- Qualitative or literature-based research analysis — use research/analysis
- Data has not been cleaned yet — use data-cleaning first
- You need charts or visual output — use visualization
- You need a formatted report — use reporting
Instructions
- Load the dataset specified in
$ARGUMENTS. Confirm the file exists and inspect its schema (columns, types, row count).
- Compute descriptive statistics:
- Central tendency: mean, median, mode for numeric columns
- Spread: standard deviation, min, max, percentiles
- Counts and frequencies for categorical columns
- Identify patterns:
- Compute correlations between numeric variables
- Detect trends over time-series data (if timestamps present)
- Flag anomalies or outliers (values beyond 2-3 standard deviations)
- Test hypotheses if specified in
$ARGUMENTS:
- State the hypothesis clearly
- Select appropriate test (t-test, chi-square, regression)
- Report results with confidence levels
- Segment the data if grouping criteria are specified:
- Group by relevant dimensions
- Compare metrics across segments
- Summarize findings with supporting evidence (specific numbers, not vague claims).
Output
## Analysis Results
### Dataset Overview
- Records: [count]
- Period: [date range if applicable]
- Key columns: [list]
### Key Metrics
| Metric | Value | Notes |
|--------|-------|-------|
| ... | ... | ... |
### Findings
1. **[Finding]**: [Evidence with numbers]
2. **[Finding]**: [Evidence with numbers]
### Correlations
- [variable A] <-> [variable B]: [coefficient]
### Anomalies
- [description with specific values]