month-heads-up
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources, creating a vulnerability surface for instructions potentially embedded in the financial records it processes.
- Ingestion points: Data is retrieved from accounting systems (
~~бухгалтерия) and payment gateways (~~платежи) during the snapshot and obligation steps. - Boundary markers: The skill does not provide the agent with specific delimiters or "ignore instructions" warnings to wrap around the ingested data.
- Capability inventory: The skill's configuration allows the use of
BashandWebFetchtools, which represent high-value targets for potential injection attacks. - Sanitization: There are no instructions for the validation, filtering, or escaping of data fetched from the external connectors.
Audit Metadata