The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process massive tabular datasets from various external sources, creating a potential surface for indirect prompt injection if the agent is instructed to reason over untrusted data contents.
Ingestion points: Data is loaded from local and remote files (CSV, HDF5, Parquet, Arrow) and cloud providers (S3, GCS, Azure) as documented in references/io_operations.md.
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to ignore potentially malicious instructions embedded within data columns.
Capability inventory: The skill includes file writing capabilities (export_hdf5), network access (cloud storage protocols, web sockets), and server management (vaex-server) as described in references/io_operations.md.
Sanitization: There are no explicit sanitization routines or validation checks described for the data contents prior to processing.
[DYNAMIC_EXECUTION]: The skill documents the use of Numba's Just-In-Time (JIT) compilation to optimize user-defined functions at runtime. In references/performance.md, examples show the use of @numba.jit to compile and execute machine code from Python functions for enhanced performance.
[EXTERNAL_DOWNLOADS]: The skill provides patterns for downloading and streaming data from established cloud services including Amazon S3, Google Cloud Storage, and Azure Blob Storage, as well as connecting to remote Vaex servers via WebSockets.
[COMMAND_EXECUTION]: The documentation in references/io_operations.md includes shell commands for launching the vaex-server to expose datasets over the network.

vaex