The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install vaex and its sub-packages (vaex-core, vaex-viz, vaex-hdf5, vaex-ml) using uv pip from PyPI, an established package registry.
[EXTERNAL_DOWNLOADS]: Documents optional cloud storage support requiring standard libraries such as s3fs, gcsfs, and adlfs from well-known providers (AWS, Google Cloud, Azure).
[EXTERNAL_DOWNLOADS]: Includes examples for integrating standard scientific computing and machine learning libraries including pandas, numpy, scikit-learn, xgboost, lightgbm, catboost, tensorflow, keras, numba, sqlalchemy, matplotlib, seaborn, and plotly.
[COMMAND_EXECUTION]: Provides legitimate commands for package installation and for starting a vaex-server to enable remote data access, which are documented features of the library.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large datasets from various file formats (CSV, HDF5, Arrow, Parquet). While this constitutes a data ingestion surface, it is the primary intended purpose of the tool and no malicious instructions were detected.
Ingestion points: Data loading via vaex.open(), vaex.from_csv(), and cloud storage paths in SKILL.md and references/io_operations.md.
Boundary markers: None explicitly defined in the prompts, as the skill focuses on code library usage.
Capability inventory: File reading and writing (export_hdf5), bash execution for installation, network access for cloud storage and web sockets, and dynamic execution via Numba JIT compilation.
Sanitization: Standard library handling of data formats; the agent is expected to follow secure coding practices when processing untrusted files.

vaex