pandera-validation

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from external DataFrames, creating a surface for indirect prompt injection attacks where data content could influence the agent's behavior.
  • Ingestion points: The validate_with_errors and infer_and_export_schema functions in scripts/schemas.py take pd.DataFrame objects as input, which are often loaded from external files or APIs.
  • Boundary markers: There are no explicit delimiters or specific instructions for the agent to ignore natural language instructions embedded within the DataFrame content.
  • Capability inventory: The skill is granted access to the Bash tool in SKILL.md, which could be exploited if an injection leads the agent to execute shell commands.
  • Sanitization: The skill performs structural and type-level validation via Pandera but does not include sanitization or filtering for potential prompt injection strings within string-based columns.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 11:49 AM
Security Audit — agent-trust-hub — pandera-validation