skills.rest
AnderHonorato
AnderHonoratoCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

exploratory-data-analysis

Inspect scientific data files with bounded, redacted local EDA reports.

Updated Aug 28, 2026
Search Tags:#scientific data#exploratory data analysis#data leakage#hdf5#fastq#missingness audit#csv profiling
One-click install
npx skills add https://github.com/AnderHonorato/Mem-rias-IA---Infinity --skill exploratory-data-analysis-anderhonorato
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: exploratory-data-analysis
Source: https://github.com/AnderHonorato/Mem-rias-IA---Infinity/tree/main/Manus/Skills/snapshots/kdense-exploratory-data-analysis
Command: npx skills add https://github.com/AnderHonorato/Mem-rias-IA---Infinity --skill exploratory-data-analysis-anderhonorato

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy, h5py, biopython, pillow, tifffile, pandas, polars, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Scientific datasets arrive in many formats and often contain sensitive identifiers, malformed structures, or hidden quality issues. This Skill performs safe, bounded exploratory data analysis on local files without executing embedded content, leaking raw values, or making unsupported statistical claims. ## Core Features & Use Cases - Bounded format inspection: Profile CSV/TSV/JSON with the Python standard library, and optionally inspect NumPy NPY/NPZ, HDF5 metadata, FASTA/FASTQ sequences, and PNG/JPEG/TIFF image metadata with pinned dependencies. - Data quality audits: Run missingness, leakage, distribution, outlier, and transformation-sensitivity analyses that flag issues without modifying raw data. - Fail-closed safety: Unknown formats are rejected, reference-only formats route to domain tooling, and all identifiers are tokenized unless explicitly revealed. - Use Case: A researcher receives a CSV of clinical measurements and runs the missingness/leakage audit with group, entity, split, and time columns to detect split overlap and missingness gaps before any modeling. ## Quick Start Ask the assistant to run a bounded exploratory analysis and missingness audit on your approved local CSV file using this skill.

Frequently Asked Questions about exploratory-data-analysis

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

FAQPage Schema
How do I run exploratory data analysis on a CSV file locally?▼

Run the bundled eda_analyzer.py script with the file path, an approved --root directory, and an optional --max-rows limit. It produces a bounded JSON report with schema, aggregate profile, and missingness statistics without printing raw row values.

What file formats does this EDA tool support?▼

Automated analysis covers CSV, TSV, and strict JSON with the standard library, plus optional NumPy NPY/NPZ, HDF5 metadata, FASTA/FASTQ, and PNG/JPEG/TIFF metadata with pinned dependencies. Formats like Parquet, BAM, DICOM, or mzML are reference-only, and unknown formats fail closed.

How do I detect data leakage between train and test splits?▼

Use missingness_leakage_audit.py with --entity-column, --group-column, --split-column, and --time-column arguments. It reports entities or groups appearing in multiple splits, duplicate rows across splits, and temporal interval overlaps as diagnostic flags.

Does the tool modify or clean my raw data?▼

No. All scripts are read-only against inputs and never delete outliers, impute missing values, normalize, or transform data. Outlier fences and transformation comparisons are sensitivity diagnostics only, and derived outputs are written separately with atomic writes.

Why does inspection fail on my HDF5 or NPZ file?▼

Inspection fails when optional dependencies are missing, signatures mismatch the declared suffix, or safety limits are exceeded. NPZ files are rejected for encryption, path traversal, high compression ratios, or object arrays requiring pickle, which is never allowed.

Can I see actual column names and identifiers in the report?▼

By default all identifiers are replaced with deterministic pseudonymous tokens. Passing --reveal-identifiers shows bounded sanitized basenames and field names, but never full paths, row values, sequence titles, or metadata attribute values.

Related Skills

Explore More Data & Analytics Skills→
kaivyykaivyy
20

download-manager

Download files from URLs into workspace folders with optional SHA-256 verification.

Community
Intermediate
pur3v4d3rpur3v4d3r
6

scikit-learn

Train and evaluate classical machine learning models with scikit-learn.

Community
Advanced
hellonishhellonish
1

statistical_analysis

Compute mean, median, range, and count for numeric columns in tabular data.

Community
Intermediate
DrugClawDrugClaw
126

medical-data-tools

Inspect DICOM metadata, analyze biosignals with NeuroKit2, and profile clinical cohort tables.

Official
Intermediate