Clickhouse Best Practices

This is a structured review framework for ClickHouse work, organized as 31 rules you check against when designing schemas, writing queries, or building data pipelines. It handles the things that trip people up: primary key column ordering (low to high cardinality, which is counterintuitive), when to actually use partitioning (data lifecycle, not query performance), and avoiding mutations in favor of ReplacingMergeTree. The agent connectivity section is solid if you're building LLM tools that query ClickHouse directly. It includes a discovery workflow and safety checks. The output format gives you a checklist structure with compliant/violation status per rule, which makes reviews faster than digging through docs every time.

Install to Claude Code

npx -y skills add clickhouse/agent-skills --skill clickhouse-best-practices --agent claude-code