Разведочный анализ научных файлов перед моделированием

★ 7.0 · data

exploratory-data-analysis is a Claude Code skill that performs bounded, deterministic exploratory analysis of authorized local scientific files before modeling or confirmatory inference. It natively supports CSV/TSV, strict JSON, NumPy (.npy/.npz), HDF5 metadata, FASTA/FASTQ streaming, PNG/JPEG container metadata, and TIFF/OME-TIFF page metadata — each with explicit read depth and safety constraints; formats such as PDB, SAM/BAM, DICOM, Parquet, and others are reference-only and require separately validated domain tooling. Reports cover schema profiling, missingness and leakage audits, outlier detection, and transformation sensitivity, without exposing raw rows, identifiers, or full paths. Researchers in bioinformatics, proteomics, microscopy, and related fields will find it useful for a rigorous, network-free data quality audit at the start of an analysis pipeline. The bundled core requires Python 3.11+; the full optional inspector set requires Python 3.12+ and uv.