The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the vaex meta-package and its legitimate sub-packages (vaex-core, vaex-viz, vaex-hdf5, vaex-ml) as well as essential cloud storage handlers (s3fs, gcsfs, adlfs) and standard data science libraries (pandas, numpy, pyarrow, scikit-learn, xgboost, lightgbm, catboost, tensorflow). All mentioned packages are industry-standard and widely used in the scientific computing ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external datasets from various formats and locations.
Ingestion points: Uses vaex.open() and vaex.from_csv() to read data from local files (HDF5, CSV, Arrow, Parquet) and remote cloud storage services (AWS S3, Google Cloud Storage, Azure Blob Storage) as documented in references/core_dataframes.md and references/io_operations.md.
Boundary markers: No explicit instruction-level boundary markers are defined for the content of the datasets themselves.
Capability inventory: The skill possesses the capability to execute shell commands (via the Bash tool), write or edit files (Write, Edit tools), and read arbitrary files on the system.
Sanitization: There is no specific mention of sanitization or instruction-filtering for text data read from external files before it is processed by the agent.
[DYNAMIC_EXECUTION]: The reference documentation in references/performance.md describes the use of Numba's Just-In-Time (JIT) compilation (@numba.jit) to optimize custom mathematical calculations on DataFrames. This is a standard performance optimization practice in Python data science and does not involve executing untrusted remote code.

vaex