csv-analyzing

Pass

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: SAFE
Full Analysis
  • Indirect Prompt Injection (LOW): The skill defines patterns for ingesting data from 'data.csv', creating a potential surface for instructions embedded in external files.
  • Ingestion points: data.csv (referenced in multiple SQL snippets).
  • Boundary markers: Absent; the templates do not include delimiters or instructions to ignore data-embedded commands.
  • Capability inventory: SQL query execution via DuckDB.
  • Sanitization: Absent; there is no validation or escaping of CSV content mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 19, 2026, 08:05 AM