node

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 analyzes StarRocks audit logs (fe.audit.log) through a provided Python script (scripts/analyze_logs.py). These logs contain raw SQL statements executed by users. An adversary could inject malicious instructions into SQL comments or query strings, which might influence the agent's behavior when it parses and acts upon the log data.\n
  • Ingestion points: fe.audit.log, fe.log, and be.WARNING are read using cat, grep, and the custom analysis script.\n
  • Boundary markers: Not present. There are no delimiters or specific instructions to the agent to ignore potentially malicious content within the logs.\n
  • Capability inventory: The skill uses curl, python3, jstack, find, grep, and tail for system interaction.\n
  • Sanitization: The analyze_logs.py script performs field extraction via regex but does not sanitize the output to prevent downstream LLM injection.\n- [COMMAND_EXECUTION]: The skill relies on executing various system utilities (top, jstack, find, grep, tail, cat, python3) to perform diagnostics. While appropriate for the skill's purpose, this represents a broad execution capability that requires oversight.\n- [DATA_EXFILTRATION]: Instructions include using curl to access internal metrics endpoints like http://<BE_IP>:<BE_HTTP_PORT>/mem_tracker. Although used here for legitimate local diagnostics, the capability to perform network requests while accessing sensitive logs poses a potential exfiltration risk if the tool is redirected.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 08:38 PM