qonto-prescription-guard
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from transaction labels, references, and invoice descriptions to identify interruptive acts for limitation periods. Because these fields are often populated by third-party debtors, they represent a vector for indirect prompt injection.
- Ingestion points: The skill ingests data via
list_transactions(labels, references) andlist_client_invoices(invoice numbers, descriptions) to perform heuristic matching. - Capability inventory: The skill utilizes
ReadandWritetools alongside the Qonto MCP tools to manage data and generate HTML timeline artifacts. - Sanitization: While the skill instructs the agent to mask IBANs (showing only the last 4 digits), it lacks specific sanitization or escaping mechanisms for the natural language content within transaction labels.
- Boundary markers: No explicit delimiters or instructions are used to signal the agent to ignore embedded instructions within the processed financial data.
Audit Metadata