dataset-insights

Profile datasets and surface ranked, reviewer-scored insights to Markdown files.

1|Updated Apr 15, 2026
One-click install
npx skills add https://github.com/gp201/agent-skills --skill dataset-insights
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: dataset-insights
Source: https://github.com/gp201/agent-skills/tree/main/skills/dataset-insights
Command: npx skills add https://github.com/gp201/agent-skills --skill dataset-insights

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Profile a dataset and surface a small, ranked list of insights, each backed by data and a reviewer score.

Core Features & Use Cases

  • Profile datasets (parquet/csv/duckdb/sql) to extract schema, univariate statistics, and red flags.
  • Generate up to 7 insights, each with evidence, justification, caveats, and confidence.
  • Reviewer pass against INSIGHT_RUBRIC.md scores and suggests KEEP / STRENGTHEN / CUT.

Quick Start

Provide the dataset path and an output slug; the tool will profile the data and produce analysis/<slug>/.

Frequently Asked Questions about dataset-insights

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I profile a dataset to surface ranked insights automatically?

To profile a dataset and surface ranked insights, provide the dataset path and an output slug. The tool extracts schema, univariate statistics, and red flags, then generates up to seven scored insights with evidence and confidence levels.

What is the best way to triage a new dataset for open-ended discovery?

The best way to triage a new dataset is running an automated profiling pass to extract schema and statistics, followed by generating a small, ranked list of reviewer-scored insights to guide initial discovery.

Does dataset profiling require DuckDB or Polars to extract schema and statistics?

Yes, dataset profiling requires either Polars or DuckDB to extract schema, univariate statistics, and red flags. The tool uses these frameworks to process parquet, csv, duckdb, or sql data sources.

How do I review generated dataset insights against a scoring rubric?

To review generated dataset insights, a reviewer pass evaluates each insight against the INSIGHT_RUBRIC.md scoring system. The reviewer outputs scores and suggests whether to KEEP, STRENGTHEN, or CUT individual insights.

What files are output when profiling datasets and generating insights?

Profiling datasets and generating insights outputs a profile.md file containing schema and statistics, and an insights.md file containing the ranked, reviewer-scored insights, both saved to an analysis directory.