LABSCO SUMMARY
ClickHouse's own 11 skills split two ways — chdb embedded in Python, and a Node.js client against a live server — with best-practice and architecture guidance layered on top of both.
The chdb pair runs entirely inside Python: chdb-datastore is a pandas-compatible DataFrame API backed by ClickHouse, and chdb-sql is embedded ClickHouse SQL over local files, S3, or remote databases, both without a server. A second track of three skills — clickhouse-js-node-coding, clickhouse-js-node-rowbinary, and clickhouse-js-node-troubleshooting — covers the separate @clickhouse/client Node package end to end: writing idiomatic client code, encoding and decoding RowBinary streams, and diagnosing errors like socket hang-ups or a stale TLS certificate. Above both tracks sit clickhouse-best-practices, a compiled set of 31 schema, query, and ingestion rules ranked by impact, and clickhouse-architecture-advisor, which turns those rules into workload-specific design decisions labeled official, derived, or field.
This is for someone already writing Python or Node.js against ClickHouse or chdb who wants an agent that reaches for the right table function or client method instead of guessing. clickhouse-managed-postgres-rca narrows further still, to a team running ClickHouse's managed Postgres offering and needing an evidence-based root-cause workflow rather than a rule set.
clickhouse-best-practices — the "MUST USE" rule set the README frames as required reading before any schema or query review — doesn't spell out its requirements; the chdb pair and the three Node-client skills need chdb or the Node client already installed, and clickhouse-managed-postgres-rca needs a live ClickHouse Cloud account. Only clickhouse-architecture-advisor is ready with nothing further to set up.
WHAT'S INSIDE
Showing all 11 skills
A design consultation for a ClickHouse database, shaped around the kind of data and traffic you actually have, where every suggestion is marked as either backed by the official documentation or drawn from field experience.
Thirty-one rules for getting ClickHouse right, used to go over a table design, a query or the way data gets loaded and point out which of them are being broken and what to change instead.
Connects an OpenTelemetry collector to a Managed ClickStack service on ClickHouse Cloud, sends realistic test logs, traces and metrics through it, and confirms you can actually see that data in ClickStack.
Gets ClickHouse, a fast database for analytics, running either on your own computer for development or as a managed ClickHouse Cloud service for production, all through one command-line tool.
Runs Postgres databases through ClickHouse's command-line tool: Docker-backed copies on your own machine for development, or managed Postgres in ClickHouse Cloud, currently in beta, for production.
A faster stand-in for pandas, the Python library for working with tables of data: your code stays as it is, but it now runs on ClickHouse's engine and can pull data straight from databases and cloud storage.
Runs SQL queries over files on your disk, over cloud storage and over other databases, all from inside a Python script and with no database server to install or keep running.
Writes the server-side JavaScript that an application uses to talk to a ClickHouse database — the connection, the data going in, and the results coming back.
Generates the JavaScript that packs and unpacks ClickHouse's compact binary data format, which is worth the trouble when you are moving huge volumes of numbers rather than text.
Something has gone wrong between a Node.js app and its ClickHouse database — this works back from the symptom to the actual cause, and names the cause before handing you the fix.
A slow or overloaded Postgres database on ClickHouse Cloud, diagnosed from the outside — the monitoring numbers and slow-query statistics the service exposes are all there is to go on.
28 best practice rules for ClickHouse schema design, query optimization, and data ingestion — prioritized by impact
Workload-aware architecture decision skill for ClickHouse — ingestion strategies, partitioning, enrichment, upsert patterns, and pre-aggregation with explicit official/derived/field provenance
Troubleshooting guide for the ClickHouse Node.js client (@clickhouse/client) — socket hang-up/ECONNRESET, Keep-Alive, data type mismatches, read-only users, proxy/pathname URLs, TLS, compression, logging, and query parameters
Pandas-compatible DataStore API for chDB — same pandas API backed by ClickHouse for 10-100x faster filter/group/aggregate/join, with built-in readers for S3, MySQL, PostgreSQL, MongoDB, ClickHouse Cloud, Iceberg, and Delta Lake and cross-source DataFrame joins
Embedded ClickHouse SQL engine for Python via chDB — run analytical SQL on parquet/csv/json files, URLs, S3, and remote databases (Postgres, MySQL, MongoDB, ClickHouse Cloud, Iceberg, Delta Lake) with Session for stateful pipelines, parametrized queries, UDFs, and cross-source joins
Local and cloud workflows for running ClickHouse with clickhousectl — local development (install ClickHouse, start a server, create schemas, seed data) and ClickHouse Cloud deployment (authenticate, create a service, migrate schemas, connect an application), with a decision tree routing between them
Local and cloud workflows for running Postgres with clickhousectl — local Docker-backed Postgres for development (start, psql client, .env wiring, lifecycle) and managed ClickHouse Cloud Postgres services, beta (authentication, service creation, connections and TLS, runtime config, read replicas, failover, point-in-time restore), with a decision tree routing between them
Step-by-step workflow for wiring an OpenTelemetry collector into a Managed ClickStack service on ClickHouse Cloud — deploying a new local collector or configuring an existing one, creating a dedicated ingest user, sending synthetic telemetry, and verifying it in ClickStack
INSTALL
CHOOSE SKILLS
Run this outside your AI tool, using Node.js 22.20.0 or newer. Skills CLI 1.7.2 starts with no skills selected; prompts cancel installation without an interactive terminal. For project installation, open your project's folder first.
Install checklist
Select All, then Enter. Enter alone cannot continue with nothing selected.Additional agents. ↑/↓ moves, Space toggles selections, Enter confirms.Installing to: for your intended tool. Stop if it names a different tool.Symlink (Recommended) for shared files or Copy to all agents for separate copies; review the summary before confirming installation.Install the find-skills skill? defaults to Yes, adding another skill for all projects. Choose No if you don't want that extra installation.npx skills add clickhouse/agent-skillsSINGLE SKILL
Run this outside your AI tool, using Node.js 22.20.0 or newer. Skills CLI 1.7.2 skips skill selection; prompts cancel installation without an interactive terminal. If choosing Project when asked about installation scope, open your project's folder before running this.
Install checklist
Additional agents. ↑/↓ moves, Space toggles selections, Enter confirms.Installing to: for your intended tool. Stop if it names a different tool.Symlink (Recommended) for shared files or Copy to all agents for separate copies; review the summary before confirming installation.Install the find-skills skill? defaults to Yes, adding another skill for all projects. Choose No if you don't want that extra installation.Find the skill name in the Skills tab and put it in the command before running it. Each entry there installs independently.
npx skills add clickhouse/agent-skills --skill <YOUR_SKILL_NAME> --full-depthPLUGIN MARKETPLACE
/plugin marketplace add clickhouse/agent-skills/plugin install clickhouse-best-practices@clickhouse-agent-skills/plugin install clickhouse-architecture-advisor@clickhouse-agent-skills/plugin install clickhouse-js-node-troubleshooting@clickhouse-agent-skillsTyped inside the agent's own prompt, not in a terminal. The marketplace lists 8 plugins; three are shown. Every one installs the same way, with its own name before @clickhouse-agent-skills.