Skip to main content

DataHub Skills

Open Source Skills Registry for Data — April 2, 2026

DataHub is the first open source metadata platform to ship a skills registry that gives AI agents direct access to catalog workflows — search, lineage, enrichment, and quality. Read the announcement blog post.

DataHub Skills give your AI agent the instructions to work with your data catalog — not just the ability to call APIs, but the knowledge of how to combine those calls into real workflows. Search for trustworthy data, trace lineage, curate metadata, and monitor quality, all from inside Claude Code.

Skills vs. Tools: The MCP Server gives agents tools — discrete actions like searching a catalog or reading a schema. Skills give agents instructions — how to chain those tools into multi-step workflows with the right judgment at each step. You need both.

Prerequisites​

Step 1: Install Skills​

The fastest path is the Skills CLI:

npx skills add datahub-project/datahub-skills

Or install directly as a Claude Code plugin:

claude plugins install datahub-skills --from github:datahub-project/datahub-skills

Verify the install:

claude plugins list

You should see datahub-skills in the output.

Step 2: Connect to DataHub​

Run the datahub-setup skill to configure your connection:

/datahub-skills:datahub-setup

The skill walks you through connecting to your DataHub instance. It configures authentication and verifies connectivity so the other skills can start working immediately.

You'll need your tenant URL and a personal access token:

Connect to DataHub Cloud at <tenant>.acryl.io

Step 3: Try the Skills​

DataHub ships with five skills. Each one handles a distinct workflow, but they hand off cleanly to each other — a search can flow into a lineage trace, which can surface a quality issue, which can trigger enrichment.

Search: Find Trustworthy Data​

Use datahub-search to explore your catalog the way your best analyst would — leveraging descriptions, glossary terms, ownership, usage stats, and quality signals.

/datahub-skills:datahub-search

Find the canonical revenue table used by the Finance team

The skill doesn't just match table names. It checks descriptions, glossary terms, ownership, usage patterns, and quality signals to surface the right asset, not just the first match.

Lineage: Trace Data Flows​

Use datahub-lineage to understand where data comes from and who depends on it.

/datahub-skills:datahub-lineage

Show me everything upstream of the weekly_revenue_report dashboard

The skill traces how data flows through your stack — upstream sources, transformations, and downstream consumers. Use it to understand impact radius before making changes.

Enrich: Curate Metadata at Scale​

Use datahub-enrich to add descriptions, tags, glossary terms, owners, and more.

/datahub-skills:datahub-enrich

Add descriptions to all undocumented columns in the orders table based on the column names and existing context

This is the skill that turns your agent into a data steward — applying metadata at a scale no human team can match.

Quality: Monitor and Investigate​

Use datahub-quality to find unhealthy assets, investigate incidents, and set up monitoring.

/datahub-skills:datahub-quality

Which tables owned by the Analytics team have failing quality checks?

For DataHub Cloud customers, the skill can also create freshness, volume, and column-level assertions on your most important tables.

Putting It Together: Build an Agent Workflow​

The real power is combining skills into multi-step workflows. Here are three patterns data engineers commonly use:

Data Analytics Agent (Text-to-SQL)​

Ask a business question in plain English. The agent uses datahub-search to find the right table — checking descriptions, glossary terms, sample queries, usage stats, and quality signals — then generates and executes SQL against your warehouse.

What was our revenue by region last quarter?

The agent finds the canonical revenue table in DataHub, confirms it's the certified source, generates the SQL, and runs it.

Data Quality Agent​

Audit and monitor your data estate. The agent uses datahub-search to find important tables by usage, datahub-quality to check assertion results and open incidents, and generates a health report.

Give me a health report for all tables owned by the Finance team, broken down by quality check status

Data Steward Agent​

Apply governance at scale. The agent uses datahub-search to find target assets, datahub-enrich to apply glossary terms and descriptions, and reports on coverage gaps.

Tag all columns containing email addresses with the PII glossary term across our Snowflake datasets, then show me a coverage report

Available Skills Reference​

SkillCommandWhat it does
Setup/datahub-skills:datahub-setupConnect to your DataHub instance (Cloud or self-hosted)
Search/datahub-skills:datahub-searchFind data assets using descriptions, glossary terms, ownership, usage, and quality signals
Lineage/datahub-skills:datahub-lineageTrace upstream sources, transformations, and downstream consumers
Enrich/datahub-skills:datahub-enrichAdd descriptions, tags, glossary terms, owners, domains, and structured properties
Quality/datahub-skills:datahub-qualityFind unhealthy assets, investigate incidents, create quality assertions (Cloud)

Other Platforms​

DataHub Skills also work with Cursor, OpenAI Codex, GitHub Copilot, Gemini CLI, and Windsurf.

# Cursor
npx skills add datahub-project/datahub-skills -a cursor

# GitHub Copilot
npx skills add datahub-project/datahub-skills -a github-copilot

# OpenAI Codex
npx skills add datahub-project/datahub-skills -a codex

# Gemini CLI
npx skills add datahub-project/datahub-skills -a gemini-cli

# Windsurf
npx skills add datahub-project/datahub-skills -a windsurf

For full installation options and manual setup, see the datahub-skills repository.

What's Next​