py-data

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data sources (CSV, Parquet, JSON), which represents a potential attack surface for indirect prompt injection if those files contain malicious instructions.
  • Ingestion points: Data enters the agent context via pl.scan_csv, pl.read_parquet, pd.read_csv, and SQL queries through duckdb.sql in SKILL.md and references/examples.md.
  • Boundary markers: The provided code examples do not include explicit boundary markers or warnings to the agent regarding potential instructions embedded in processed data.
  • Capability inventory: The skill demonstrates capabilities for file writing (write_parquet, to_csv) and data output to the console.
  • Sanitization: Examples show standard data cleaning (clipping, null removal, regex for format validation), but no specific sanitization for prompt injection strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 06:31 PM