resource-isolation
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides templates for processing cluster logs (audit and system logs) using utilities like
grepandawk. This represents a standard indirect prompt injection surface where untrusted data within logs could potentially interfere with parsing logic. - Ingestion points: Accesses
fe/log/fe.audit.log,fe/log/fe.log, andbe/log/be.logfor troubleshooting. - Boundary markers: No explicit delimiters are suggested for isolating log content from the analysis context.
- Capability inventory: Includes shell command execution (
curl,grep,awk,echo) and SQL execution for cluster management. - Sanitization: Log processing commands rely on basic text matching and field splitting without advanced sanitization of the processed content.
Audit Metadata