kubesense-sql
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data originating from application logs and traces via the execute-sql tool. * Ingestion points: Results from execute-sql and validate-sql in SKILL.md. * Boundary markers: Tool output is structured as TSV with header metadata. * Capability inventory: Retrieved data is used by the agent to analyze system state and formulate follow-up actions. * Sanitization: SQL syntax is validated server-side, but the content of returned log/trace rows is not sanitized for natural language instructions.
- [DYNAMIC_EXECUTION]: The skill enables the agent to generate and run dynamic ClickHouse SQL queries. * The execute-sql and validate-sql tools allow arbitrary SQL construction. * Mitigation: The environment is restricted to read-only access on specific tables (logs, traces) and blocks dangerous functions such as url, s3, and remote.
Audit Metadata