balance

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch and process metrics from external network endpoints (StarRocks FE and BE nodes). This ingestion of data from network resources lacks explicit boundary markers or sanitization, creating a surface where malicious content returned from a compromised node could potentially influence agent behavior.
  • [COMMAND_EXECUTION]: The skill includes instructions for executing bash shell commands, such as loops iterating over network addresses, and administrative SQL commands that modify critical system configurations (e.g., ADMIN SET FRONTEND CONFIG).
  • [DATA_EXFILTRATION]: The instructions utilize curl to perform network operations targeting non-whitelisted IP addresses and placeholders for cluster nodes. While these are intended for cluster monitoring, they represent network communication with non-standard domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 08:37 PM