data-engineering-storage-remote-access-integrations-pyarrow

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWPROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection (LOW): The skill defines patterns for ingesting data from external cloud storage buckets, creating a potential surface for indirect instructions to enter the agent's context.
  • Ingestion points: The skill utilizes pq.read_table() and ds.dataset() to pull Parquet data from S3, GCS, or Azure buckets as seen in SKILL.md.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded content are present in the provided code snippets.
  • Capability inventory: The skill facilitates reading remote files using pyarrow.fs and fsspec compatibility layers.
  • Sanitization: There is no evidence of content sanitization or schema validation beyond format-specific parsing.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 08:01 AM