Skip to main content

Quantitative Analysis

Profile CSVs and compute transparent numeric summaries.

Skill metadata​

SourceBundled (installed by default)
Pathskills/research/quantitative-analysis
Version1.0.0
AuthorVibeOS
LicenseMIT
Platformslinux, macos, windows
Tagsresearch, analysis, csv, statistics, execute-code
Related skillscomparative-analysis, professional-research, longform-synthesis

Reference: full SKILL.md​

info

The following is the complete skill definition that VibeOS loads when this skill is triggered. This is what the agent sees as instructions when the skill is active.

Quantitative Analysis Skill

Turn tabular data into clear numeric findings with reproducible steps. Prefer execute_code (pandas if available) or the helper scripts here.

When to Use​

  • User provides CSV/TSV/XLSX or asks for stats, trends, cohorts, sensitivity.
  • Need descriptive stats before a narrative conclusion.

Prerequisites​

  • Data file path known (or downloaded via tools).
  • execute_code and/or terminal available.
  • Do not invent rows; if a column is missing, say so.

How to Run​

  1. Profile the file (shape, dtypes, nulls, sample rows).
  2. Clean only with explicit rules (document drops/filters).
  3. Compute the metrics the question needs.
  4. Report with formulas, units, and caveats (sample size, outliers).
  5. Optional: write a small chart/table file via execute_code.

Quick profile (stdlib):

python3 scripts/profile_csv.py path/to/data.csv

Output template​

# Quantitative brief: <question>

## Dataset
- Path / URL:
- Rows × cols (after filters):
- Filters applied:

## Key metrics
| Metric | Value | Method |
|--------|-------|--------|

## Findings
1. …
2. …

## Sensitivity / caveats
- Missingness, outliers, selection bias

## Reproduce
Short `execute_code` or CLI snippet

Pitfalls​

  • Correlation ≠ causation — label associations clearly.
  • Percentages need denominators.
  • Don’t silent-drop nulls without stating count removed.
  • Currency/timezone units must stay consistent.

Verification​

  • Row counts before/after filters stated.
  • Each metric has a method note.
  • Reproduce snippet works on the same file.