dbs-chatroom-austrian
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input and interpolates it directly into the prompts for simulated sub-agents (Hayek and Mises).
- Ingestion points: User-provided topics and follow-up questions are captured via
{用户问题}and{新问题}placeholders inSKILL.md. - Boundary markers: The prompts for the sub-agents do not use delimiters (like XML tags or markdown blocks) or explicit instructions to ignore commands within the user's input.
- Capability inventory: While the sub-agents are primarily used for text generation, successful injection could lead to jailbreaking the personas or influencing the 'Claude Judge' summary.
- Sanitization: There is no evidence of input validation or escaping for the user-provided data before it is inserted into the sub-agent instructions.
Audit Metadata