Loading...
Loading...
Found 27 Skills
Read and write large cuPyNumeric arrays to HDF5 with Legate's parallel, distributed HDF5 I/O (legate.io.hdf5: to_file, from_file, from_file_batched). Use when a developer needs to save a cuPyNumeric array to an .h5/.hdf5 file, load an HDF5 dataset into a distributed cuPyNumeric array, read a large HDF5 dataset in chunks, hand arrays to an HPC pipeline as a single file, or accelerate HDF5 disk I/O with GPUDirect Storage (GDS). Do not use it for Parquet/cuDF/raw-binary or other sharded/custom layouts (see the cupynumeric-parallel-data-load skill), Zarr or object-store/S3 output, .npz or pickled archives, plain h5py without cuPyNumeric, or pure array compute such as FFT, matmul, or reductions.
Use when writing or running Nushell commands, scripts, or pipelines - via the Nushell MCP server (mcp__nushell__evaluate), via Bash (nu -c), or in .nu script files. Also use when working with structured data (JSON, YAML, TOML, CSV, Parquet, SQLite), doing ad-hoc data analysis or exploration, or when the user's shell is Nushell.
Fast in-process analytical database for SQL queries on DataFrames, CSV, Parquet, JSON files, and more. Use when user wants to perform SQL analytics on data files or Python DataFrames (pandas, Polars), run complex aggregations, joins, or window functions, or query external data sources without loading into memory. Best for analytical workloads, OLAP queries, and data exploration.
Explore and query any dataset annotated with a Frictionless Data Package descriptor (datapackage.json). Use this skill whenever a user wants to discover what tables or resources a dataset contains, look up column names and descriptions, surface usage warnings embedded in metadata, or understand how to load data from Parquet files, DuckDB or SQLite databases, or CSV files described by a datapackage.json. Also use when the user has a datapackage.json and wants to know what's in it, how to query it efficiently, or how to connect its metadata to actual data files. Pairs well with dataset-specific skills (like `pudl`) that layer domain knowledge on top.
Profile a new tabular dataset before modeling. Find target leakage, missing data patterns, high-cardinality categoricals, near-constant features, redundant pairs, and non-linear relationships that Pearson correlation misses. Use whenever the user hands you a CSV or parquet and asks "what should I do with this?" Always run this skill before training any model on data you haven't seen before.
Explore and query data on S3, Cloudflare R2, GCS, MinIO, or any S3-compatible storage. Use when the user mentions an s3://, r2://, gs://, or gcs:// URL, asks "what's in this bucket", wants to list remote files, preview remote Parquet/CSV/JSON, or query data on object storage without downloading it. Also triggers when the user wants to know the size, schema, or row count of remote datasets.
Scalable data processing for ML workloads. Streaming execution across CPU/GPU, supports Parquet/CSV/JSON/images. Integrates with Ray Train, PyTorch, TensorFlow. Scales from single machine to 100s of nodes. Use for batch inference, data preprocessing, multi-modal data loading, or distributed ETL pipelines.
Performs gap analysis on NVIDIA TAO Visual ChangeNet (VCN) Classify experiments by invoking the data-services container (`tao_toolkit.data_services` from `versions.yaml`) directly via `docker run … gap_analysis vcn_aoi …` — picks the optimal decision threshold, ranks per-sample weakness, and emits a top-K weakest parquet expanded per-lighting for downstream augmentation. Use when analyzing VCN classification failures, picking SDA augmentation targets, auditing PASS/NO_PASS boundary cases, or running DEFT gap analysis on an AOI ChangeNet model.
This skill provides guidance for merging data from multiple heterogeneous sources (JSON, CSV, Parquet, XML, etc.) into a unified dataset. Use this skill when tasks involve combining records from different file formats, applying field mappings, resolving conflicts based on priority rules, or generating merged outputs with conflict reports. Applicable to ETL pipelines, data consolidation, and record deduplication scenarios.
Analyze datasets to discover patterns, anomalies, and relationships. Use when exploring data files, generating statistical summaries, checking data quality, or creating visualizations. Supports CSV, Excel, JSON, Parquet, and more.
Convert files between 140+ formats using the ConversionTools MCP server. Use when the user needs to convert documents (Word, PDF, Excel, PowerPoint), data formats (JSON, CSV, XML, YAML, Parquet), images (PNG, JPG, WebP, AVIF, HEIC, JXL, SVG), audio (MP3, WAV, FLAC), video (MOV, MKV, AVI to MP4), e-books (EPUB, MOBI, AZW), OCR text extraction, AI-powered data extraction, AI text-to-speech (TTS), AI speech-to-text transcription (STT), subtitle conversion (SRT, VTT, ASS), or website screenshots.
Runs the DEFT embed-then-mine workflow for VCN AOI iterations — embeds the gap-analysis target parquet, embeds a source pool, and mines nearest-neighbour source images for downstream augmentation. Use as the immediate next step after `tao-route-visual-changenet-samples` when expanding a real-image augmentation queue from the mining subset.