The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides the ability to ingest and process large datasets from external files and cloud storage, which is a standard attack surface for data processing tools.
Ingestion points: Data loading methods for CSV, HDF5, Parquet, and Arrow files, as well as cloud-based URIs (S3, GCS, Azure), are described in references/core_dataframes.md and references/io_operations.md.
Boundary markers: The skill does not implement custom boundary markers or instructions to ignore embedded instructions within data, relying on the agent's core guardrails.
Capability inventory: The skill allows for local file writing (exports), network-based I/O to cloud services, and performance-oriented runtime code generation (JIT).
Sanitization: Relies on the standard data validation and parsing provided by the Vaex library.
[DYNAMIC_EXECUTION]: The skill documents the use of Just-In-Time (JIT) compilation and custom function registration for performance optimization of arithmetic expressions.
Evidence: references/performance.md describes using .jit_numba() and .jit_pythran() to compile expressions at runtime, and the @vaex.register_function() decorator to integrate Python logic into the execution backend.
[EXTERNAL_DOWNLOADS]: The documentation references the installation of standard, well-known third-party libraries for extended functionality.
Evidence: references/io_operations.md mentions installing the s3fs, gcsfs, and adlfs packages via pip for cloud storage connectivity.

alterlab-vaex