Skip to content
Skills Library
Data & analytics · Analysis

explore

/explore

Investigate data structure, quality, relationships, or model errors with question-led exploratory analysis. Use when observations need explanation before deciding on collection, features, models, or scope; inferential claims belong to statistics.

Repository
1ml ↗
Licence
MIT
Last repository change
04.11.2026

Assessment and security

90/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content25 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 04.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Investigate data structure, quality, relationships, or model errors with question-led exploratory analysis. Use when observations need explanation before deciding on collection, features, models, or scope; inferential claims belong to statistics.

Original author description

Investigate data structure, quality, relationships, or model errors with question-led exploratory analysis. Use when observations need explanation before deciding on collection, features, models, or scope; inferential claims belong to statistics.

How it works

  1. Investigate
  2. Record and route
Original SKILL.md headings · 30 lines, 475 words
SKILL.md preview
---
name: explore
description: Investigate data structure, quality, relationships, or model errors with question-led exploratory analysis. Use when observations need explanation before deciding on collection, features, models, or scope; inferential claims belong to statistics.
---

# Explore to resolve a question

Identify the decision this analysis should inform and the evidence population it may inspect. Use [state](../../references/project-state.md) to reuse source/split definitions. Initial schema and integrity checks can precede splitting; adaptive feature/target/model exploration for prediction belongs to development data. Keep final evaluation reserved.

Read [question-led EDA](references/question-led-eda.md) for the relevant patterns and failure routes. Load a matching [task profile](../../references/tasks/index.md) when time, censoring, exposure, labels, or representation alter interpretation.

## Investigate

Inspect units, row/entity counts, ranges, missingness by source/time/group, category frequency, outcome support, duplicates/joins, and availability at decision time. For a runnable exploratory notebook, produce and inspect the figures needed to understand consequential distributions, missingness and relationships; printed summaries alone do not resolve visual questions. Choose relevant branches from the [visual evidence guide](references/question-led-eda.md#visual-evidence-and-decisions), rather than plotting every column. Compare distributions and error patterns by meaningful subgroups; record small support rather than interpreting unstable plots as facts.

Distinguish an observed association from a causal explanation. High skew alone does not imply a nonlinear model or a required transformation. A missingness percentage does not identify its mechanism. A visually separated cluster does not establish actionability. Inspect artifacts that could explain suspiciously good scores, including leakage and duplicated entities.

When considering correlation, covariance, skewness or kurtosis, follow [statistic selection](references/question-led-eda.md#choose-descriptive-statistics-for-the-decision). Use meaningful columns and report support; compute the statistic when it informs a decision, not merely to fill a checklist. Link observed structure to the feature-construction or no-change decision in prepare.

For model error analysis, use development predictions and actual costs to select representative ordinary and expensive errors. Preserve the denominator, failed predictions, and unlabeled cases. Generate specific hypotheses and test one plausible cause at a time. If analysis uses final-test outcomes to adapt a model, record that the test has become development evidence and request renewed evaluation.

## Record and route

Save reproducible calculations and rendered figures in the [notebook](../../references/notebook-protocol.md). Make the population/support, finding, relevant limit and resulting decision clear for each consequential analysis. Related figures can share one interpretation; labels, legends and a concise caption can supply context without a separate prose checklist. Explain what the evidence changes, including retaining the current approach. Verify that saved outputs actually contain the figures and that captions match the executed values. An import of a plotting library, a planned chart or an unfinished placeholder is not visual evidence. Separate measured facts, hypotheses, and unsupported claims.

- Source corruption, units, or ambiguous joins: return to [data](../data/SKILL.md).
- Invalid target/population or no useful action: return to [frame](../frame/SKILL.md).
- Statistical uncertainty, comparisons, or model-assumption inference: use [statistics](../statistics/SKILL.md).
- Candidate representation supported by evidence: use [prepare](../prepare/SKILL.md).
- Measured model failure or candidate comparison: use [train](../train/SKILL.md).
- Unsupported split or leaked observations: use [evaluation](../evaluation/SKILL.md) and invalidate affected results.

Complete when the investigation answers its question with limitations or identifies the exact missing evidence. An exhaustive gallery of plots is not a completion criterion.
Files in this package (2)
  • SKILL.md 4 KB
  • references/question-led-eda.md 9 KB

Versions

RepositoryScoreLicence
This page · Kiran-980/1ml90MIT
yonatangross/orchestkit100MIT
yonatangross/orchestkit100MIT
rapiddweller/archkeel100MIT
zimoliao/scholaraio100MIT
artk0de/tearags-mcp100MIT
artk0de/tearags-mcp100MIT
artk0de/tearags-mcp100MIT
artk0de/tearags-mcp100MIT

Database snapshot: 13.11.2026 · package version 6830cb0751fb (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 98exploratory-data-analysis
    Perform systematic exploratory data analysis to understand dataset structure, distributions, relationships, and anomalies before modeling. Use when a dataset is new, its quality is unknown, or the user requests open-ended profiling; use data-analysis instead for a defined…
  • 100stat-eda
    Conduct Exploratory Data Analysis (EDA) using descriptive statistics, visualizations, and data quality checks. Use this skill when the user has a dataset and needs to understand its structure, find patterns, detect anomalies, or prepare data for further analysis — even if they…
  • 100analysis-artifacts
    Use when the user requests a deep dive, exploratory analysis, or data analysis on BigQuery. Not for credential, publish, deploy, or irreversible changes.
  • 100alterlab-eda
    Exploratory data analysis (EDA) on a scientific data file — auto-detects the format, runs structure/quality/statistics checks, and writes a markdown EDA report with downstream recommendations. Use when asked to "explore", "analyze", "summarize", "profile", or "QC" a data file,…
  • 90exploratory-data-analysis
    Explore a new dataset systematically to understand its shape, quality, and signal before modeling. Use when you first get a dataset and need to know what is in it, what is wrong with it, and what is worth pursuing.
  • 93exploratory-data-analysis
    Use this skill when exploring an unfamiliar dataset, investigating data quality issues, validating data pipeline outputs before production use, reviewing SQL queries for correctness and performance, or producing a data quality scorecard for a data asset.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.