The Agent Skills Directory

[SAFE]: The skill provides documentation for the legitimate Python library vaex. All analyzed scripts and instructions reflect standard data science practices.
[DATA_EXFILTRATION]: The skill documents how to access remote data via S3, GCS, and Azure in references/io_operations.md. These are standard functionalities for big data processing and do not involve unauthorized exfiltration. The use of placeholders for credentials follows security best practices.
[DYNAMIC_EXECUTION]: The skill mentions using Numba (numba.jit) for performance optimization in custom calculations in references/performance.md. This is a common practice in high-performance data science to speed up array operations.
[INDIRECT_PROMPT_INJECTION]: The skill handles external data ingestion, representing a potential attack surface if processing untrusted user data. Ingestion points: Methods like vaex.open and vaex.from_csv in references/core_dataframes.md and cloud storage connectors in references/io_operations.md. Boundary markers: No explicit boundary markers or warnings are defined in the library usage examples. Capability inventory: Includes file writing (export_hdf5), network access (cloud storage), and dynamic code generation (Numba JIT) as documented in references/io_operations.md and references/performance.md. Sanitization: Relies on standard data validation provided by the Vaex library.

vaex