optimizing-clickhouse-and-hogql-queries
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted data from ClickHouse query logs (system.query_log, query_log_archive) and user-provided HogQL/SQL inputs, creating a surface for indirect prompt injection. 1. Ingestion points: ClickHouse system logs and user-supplied query strings. 2. Boundary markers: Absent. 3. Capability inventory: Execution of ClickHouse SQL via sync_execute, client.execute, and execute_hogql_query across multiple scripts. 4. Sanitization: Absent.
- [COMMAND_EXECUTION]: The skill facilitates the execution of powerful ClickHouse SQL commands and references the hogli CLI tool. These capabilities allow for significant system interaction that could be misused if the agent is misled by malicious data in the processed logs.
- [EXTERNAL_DOWNLOADS]: The skill references an external optimization tool, pi-autoresearch, hosted on a personal GitHub repository (github.com/davebcn87/pi-autoresearch). While it suggests the user run the tool, it introduces a dependency on unverified third-party code outside of the official vendor organization.
Audit Metadata