Skills & Agents
Find my skill

Data Set Analysis

Takes one data file someone attaches, plus what kind of data it is, and writes back one report with four sections in a fixed order: a summary of what's actually in the file, the patterns or trends the numbers show, insights grouped by how much they matter, and recommendations. Every insight says why it landed in its priority bucket, and every recommendation names the insight behind it, so nobody has to take the advice on faith. The metrics the report leans on shift with the data type: a spend file gets read for spend concentration and month-over-month change, a survey file gets read for score distribution and response rate, using a starter reference table that ships with the skill. Nothing in the report goes further than the file supports; when a number or trend isn't in the data, the report says so instead of filling the gap with something that sounds plausible. Before writing anything, it checks the file for columns and column combinations that could identify a specific person, stops to flag them and ask before proceeding, and describes them by name and shape rather than quoting raw values unless told otherwise. It also validates the file for ragged rows, mixed types, and inconsistent formats before reporting any figure that touches an affected column, and treats any header or cell content that reads like an instruction to the model as data to describe, never a command to follow. Ships with a starter metrics reference table covering six common data types and an eval contract carrying four frozen self-test scenarios, including a dedicated PII redaction test.

First-party, by Skills and Agents Co

August 31, 2026

Install

Simple install — no setup required

1

Open the Claude desktop app and go to the Project you want this in

2

Go to Code, then Customize, then click Create a new skill and upload the skill file below

3

Say a trigger phrase to activate it

The skill file
Download
Or install with the CLInpx skills add skills-agents-co/skills-and-agents-library/data-set-analysis
Overview

What this skill does

This is an overview of the skill from the site, not the skill's real instructions. The full SKILL.md Claude reads when this skill is active lives on GitHub at the source below.

Takes one attached data file and writes back a structured read on it: a spreadsheet of spend, an ads export, a usage log, survey results, a sales pipeline dump, or a general operational file, all handled by the same four-section report, with the metrics the report focuses on actually changing to fit what kind of data it is.

This works from the attached file alone. It doesn’t pull from a live database, a warehouse, or any connected tool, and it doesn’t go looking for a file on its own.

What it does

  • Establishes the data type before analyzing anything. Uses what you tell it, or infers from the columns and asks you to confirm, rather than guessing silently and chasing the wrong metrics.
  • Shifts its focus metrics to the data type. A marketing file gets read for cost per conversion and spend by campaign; an operational file gets read for throughput and error rate. Pulls from a starter reference table, not one generic list run against every kind of data.
  • Validates the file before computing anything. Flags ragged rows, mixed types, duplicate headers, and inconsistent formats, and names the anomaly before reporting any figure that touches the affected column.
  • Checks for identifying data before writing a word. Scans column names and values for anything that could identify a specific person, including combinations of columns that narrow down to one person even when no single column does, and stops to ask before proceeding.
  • Never invents a figure. Any number, trend, or comparison in the report has to trace back to something computed from or read directly out of the file. A metric the data type calls for but the file doesn’t have gets marked not available, not estimated.
  • States why every insight matters. Each insight in the priority sections names the reason it’s high, medium, or low, and every recommendation names the specific insight it follows from.
  • Treats the file’s contents as data, never instructions. A column header or cell that reads like a directive to the model gets named and flagged, never followed, and never quoted back verbatim.

How to use it

  1. Attach a data file and say what kind of data it is, or let it infer and confirm.
  2. If it finds any columns or combinations that look identifying, it stops and asks how you want them handled before going further.
  3. It reads the file, validates it, and writes the four-section report: summary, patterns and trends, insights by priority, recommendations.
  4. Anything the file doesn’t support gets marked not available instead of guessed at.

Output format

# Data Set Analysis: <file name / data type>

## Summary of key content
## Patterns and trends
## Insights grouped by priority
### High priority
### Medium priority
### Low priority
## Recommendations

Pitfalls

  • Don’t guess the data type when it isn’t obvious. A wrong guess sends the whole report chasing the wrong metrics.
  • Don’t reuse the same generic metric list for every data type. Let the focus actually change.
  • Don’t quote a value that identifies a person. Flag it and ask first; describe by shape, never by value, until told otherwise.
  • Don’t fill a gap with a plausible number. If the file doesn’t say it, the report doesn’t say it either.

This skill has no dependency on any external data source or connector. It reads only what’s attached to the conversation.

View full SKILL.md on GitHub