vaex

Pass

Audited by Gen Agent Trust Hub on Jun 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data, which creates a surface for indirect prompt injection.
  • Ingestion points: Data is loaded from local or remote files via vaex.open() and vaex.from_csv() as seen in references/io_operations.md and references/core_dataframes.md.
  • Capability inventory: The skill can write files (export_hdf5), train ML models, and perform complex data transformations across various scripts.
  • Boundary markers: No delimiters or instructions are present to prevent the agent from obeying instructions embedded within the processed data.
  • Sanitization: The documentation does not provide methods for validating or sanitizing input data before processing.
  • [COMMAND_EXECUTION]: The skill uses Numba JIT (@numba.jit) to compile and execute custom Python logic at runtime for performance, as documented in references/performance.md. It also references the vaex-server CLI tool in references/io_operations.md.
  • [EXTERNAL_DOWNLOADS]: The documentation includes patterns for accessing data from well-known cloud storage services, including Amazon S3, Google Cloud Storage, and Azure Blob Storage, using standard integration libraries.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 20, 2026, 12:54 PM
Security Audit — agent-trust-hub — vaex