This package contains the modular altinity-expert-clickhouse-* ClickHouse diagnostic skills and a dedicated test suite.
skills/— allaltinity-expert-clickhouse-*skill definitions (each has its ownSKILL.md). The...-overviewskill is the router and suggests which specialist skill(s) to run next.tests/— test suite and scenarios. Do not modify structure without updating prompts/runner.../releases/— built zip packages for distribution.
- Run a quick overview of ClickHouse server health
- Based on findings, run one specialist skill (e.g. memory, merges, replication)
- Full server analysis (runs all skills; can take a long time):
Run a comprehensive ClickHouse server analysis and write a single Markdown report to `report.md`.
Workflow:
1) Run `altinity-expert-clickhouse-overview` first and capture the key facts (hostName/version, uptime, disk/memory summary, system.errors, system.*_log activity, system.warnings, log TTL status).
2) Then run ALL specialist skills one-by-one and merge findings into the same `report.md`:
- altinity-expert-clickhouse-metrics
- altinity-expert-clickhouse-logs
- altinity-expert-clickhouse-memory
- altinity-expert-clickhouse-storage
- altinity-expert-clickhouse-caches
- altinity-expert-clickhouse-merges
- altinity-expert-clickhouse-ingestion
- altinity-expert-clickhouse-reporting
- altinity-expert-clickhouse-replication
- altinity-expert-clickhouse-mutations
- altinity-expert-clickhouse-dictionaries
- altinity-expert-clickhouse-schema
Report requirements:
- Use headings per module, and end with a single prioritized action plan (Top 10), each with severity and “why”.
- Include the most important supporting queries (and their results) inline or in appendices.
- If any diagnostic query fails, record it in an “Appendix: Failed Queries” section with the error and the module name.
- Make sure the report clearly states the timeframe used for logs (last 24h unless otherwise specified).
From altinity-expert-clickhouse/tests:
make help
# Default: starts a local ClickHouse in Docker (CLICKHOUSE_VERSION=25.8)
make validate
# Run a single skill test
make test-overview
# Or run everything
make test-allNotes:
- The test runner uses
tests/runner/lib/common.shhelpers andCLICKHOUSE_*env vars for connectivity. - Provider-specific targets exist in
tests/Makefile(Codex is the default; Claude is supported; Gemini is a stub). - Reports are written under
tests/reports/<skill-name>/. - For more details (Docker, version switching, SQL-only runs), see
tests/TESTING.md.
The test suite still supports pointing at any reachable ClickHouse server:
export CLICKHOUSE_HOST=<host>
export CLICKHOUSE_PORT=9000
export CLICKHOUSE_USER=default
export CLICKHOUSE_PASSWORD=... # if set
export CLICKHOUSE_SECURE=0 # or 1 if native TLS
# Disable docker-managed ClickHouse
make validate USE_DOCKER=0
# Example: run a single test
make test-overview USE_DOCKER=0From repo root (/Users/bvt/work/Skills):
makeOutputs:
releases/<skill-name>.zip
To list detected skills:
make list