altinity-expert-clickhouse-index-analysis

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill extracts and analyzes historical queries from the system.query_log table, which can include attacker-controlled input if they have access to execute queries on the ClickHouse instance.
  • Ingestion points: Data is pulled from system.query_log in both SKILL.md and checks.sql.
  • Boundary markers: There are no explicit delimiters or instructions to the agent to treat the data as untrusted or to ignore embedded instructions within the captured SQL.
  • Capability inventory: The skill facilitates the execution of diagnostic SQL and analysis of the results to recommend schema changes.
  • Sanitization: No sanitization is performed on the results of the extractAll or normalizeQuery functions before presenting them to the LLM.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:54 PM