dbt-labs dbt-agent-skills
COMMUNITYLABSCO SUMMARY
The dbt plugin covers day-to-day analytics engineering: building and modifying models, writing unit tests, building Semantic Layer metrics with MetricFlow, answering business questions against that layer, judging whether a model change breaks dbt Mesh consumers, triaging a failed Cloud job, setting up the dbt MCP server, looking up dbt docs, and running the CLI itself — nine skills in the README's own table. A separate dbt-migration plugin holds two more, labeled explicitly as one-off: moving a project across data platforms, and moving it from dbt Core to the Fusion engine. Outside those two tables, three further live skills exist that the README's install instructions never mention: auditing-skills (reviewing other published skills for security or quality issues, unrelated to dbt itself), creating-mermaid-dbt-dag (Mermaid lineage diagrams from the manifest or MCP tools), and using-dbt-state (dbt's server-backed run-skipping and cloning feature). Since the README's plugin-install commands are the only documented way in, it is not written whether installing dbt or dbt-migration gets you these three at all.
This is for a team already running dbt Core, dbt Cloud, or the Fusion engine that wants an agent to write and test models, work the Semantic Layer, or diagnose a broken job without leaving the terminal. The two migration skills matter only for the week a team is actually switching platform or engine, by the README's own description, and someone without an existing dbt project gets little from most of the rest — fetching-dbt-docs and configuring-dbt-mcp-server are the two the README itself says don't need one.
READ THE FULL ANALYSIS
Five of the 14 skills depend on a local tool being present, two need an account key, two are marked unverified in our data, and one talks to a running MCP service — only four carry no dependency at all. That spread, and the three undocumented skills, are both read from the live dbt-agent-skills tree rather than the README's own account of itself.
ALSO IN THIS PACKAGE
dbt
Skills for analytics engineering with dbt — building models, writing tests, querying the semantic layer, troubleshooting jobs, and more.
dbt-migration
Skills for migrating dbt projects — moving from dbt Core to dbt v2 or across data platforms.
dbt-extras
Miscellaneous skills for dbt.
WHAT'S INSIDE
13 showing · 13 totalNothing else to set up — install it and go.
adding-dbt-unit-test
Feed a piece of dbt transformation logic a handful of made-up rows, say what should come out the other end, and dbt will refuse to build it if that ever stops being true.
answering-natural-language-questions-with-dbt
Ask a plain question — how much did we sell last month, who are our biggest customers — and get a number back, using the metric definitions your team already agreed and writing the query itself where they don't cover it.
auditing-skills
Once you publish a skill, outside scanners grade it for security and quality; this reads those verdicts, works out the handful of real causes behind them, and fixes each one everywhere it appears.
building-dbt-semantic-layer
Writing down what your company means by revenue, or by an active customer, once and in one place on top of the tables you already have — so that everyone who asks gets the same number back.
configuring-dbt-mcp-server
The wiring between your AI coding tool and dbt: which of the two connectors to run, which one needs a paid account, and which powers it hands the tool over your project.
creating-mermaid-dbt-dag
Draws the picture of where one table's data comes from and what depends on it further down, as a diagram you can paste straight into a document.
fetching-dbt-docs
dbt's documentation site will hand back any page as plain text if you ask it the right way; this finds the page you actually need and fetches it in that form.
migrating-dbt-project-across-platforms
A step-by-step move of a dbt project onto a different database system: fixing everything the new one refuses to accept, and stopping only when the same numbers still come out.
running-dbt-commands
Three different programs are all called dbt, and the one you type is not always the one you meant; this picks the right executable and the right command for what you are actually trying to do.
troubleshooting-dbt-job-errors
A scheduled dbt job failed overnight and the error message does not say why; this is the discipline for finding out — and the rule that you never make a failing test pass before you understand it.
using-dbt-for-analytics-engineering
The everyday dbt work — building a table out of SQL, extending one, untangling one — done the way a software engineer would: nothing written twice, tests alongside, results checked before anything ships.
using-dbt-state
dbt State is the paid feature that asks a server, before each table is rebuilt, whether it could simply be reused instead; this covers turning it on, paying for it, and living with it.
working-with-dbt-mesh
Other people's models and dashboards depend on yours, so renaming a column quietly breaks them; this is how to make that change with a deprecation window instead, and how separate projects safely borrow each other's tables.
HOW TO GET IT
npx skills add dbt-labs/dbt-agent-skillsnpx skills add dbt-labs/dbt-agent-skills --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.
/plugin marketplace add dbt-labs/dbt-agent-skills/plugin install dbt@dbt-agent-marketplace/plugin install dbt-migration@dbt-agent-marketplace/plugin install dbt-extras@dbt-agent-marketplaceTyped inside the agent's own prompt, not in a terminal. The marketplace is called dbt-agent-marketplace, which is the part after the @.