swaps
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from CSV files which could contain malicious instructions designed to influence the agent's behavior.
- Ingestion points: The
scripts/swaps_validation.pyscript reads user-provided CSV data viapd.read_csv(arguments.input_csv). - Boundary markers: No explicit boundary markers or instructions to ignore embedded content are provided in the prompt interpolation logic.
- Capability inventory: The skill allows for file writing (
scripts/swaps_validation.pywriting to--output), and theSKILL.mdinstructions encourage the agent to execute Python scripts and read reference documentation. - Sanitization: The script performs type conversion using
pd.to_numeric(errors='coerce')andpd.to_datetime(errors='coerce'), which provides basic data validation but does not sanitize against prompt-based attacks.
Audit Metadata