matematic-expert-panel
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of a 'case' description provided by the user during the setup phase. This data is then used to prompt several AI personas (Compliance Officer, IT Security, etc.) for analysis. Without explicit boundary markers or sanitization, a malicious case description could attempt to override the persona instructions or influence the final decision matrix output.
- Ingestion points: The 'Faza 1
- Setup + case framing' section in
SKILL.mddescribes how the agent should frame a case for analysis. - Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions embedded within the provided case content.
- Capability inventory: The skill utilizes
ReadandWritetools to process information and generate the 'Raport pozegnal' (Final Report) output. - Sanitization: There is no evidence of sanitization, filtering, or escaping logic applied to the user-supplied case text before it is interpolated into the persona prompts.
Audit Metadata