Plain filesEach skill is a folder with a SKILL.md, copied to ~/.claude/skills. Claude loads it when your request matches.
No MCP serversClaude calls Fabric with the Azure CLI and REST APIs (az rest, curl, jq), plus sqlcmd for SQL.
Only what you pickChoose from 23 skills below and copy one command. Removing a skill later just means deleting its folder.
1

Pick your skills

Browse by area and click a skill to see what it does and what to ask it. Press + Add on each one you want. Your install command is built in step 2.

PySpark, Spark SQL, notebook runs, Materialized Lake Views, medallion design.

spark-cli

The notebook skill. Writes cell code (PySpark, Scala, SparkR, %%sql, %%configure, notebookutils), runs notebooks by name, opens Livy sessions for ad-hoc analysis, diagnoses failed or slow Spark runs, and manages Materialized Lake Views.

Use cases

  1. Write a notebook

    Create a notebook in workspace 'Sales-Dev' that reads the raw orders CSVs from lh_bronze/Files/orders, dedupes by order_id and writes a Delta table silver.orders.

  2. Debug a failed run

    The notebook nb_silver_orders failed last night with an out-of-memory error. Find the run, tell me why it failed and fix the Spark config.

  3. Ad-hoc analysis

    Open a Livy session on lh_silver and show me daily order counts for the last 30 days, split by country.

  4. Materialized Lake View

    Create a Materialized Lake View gold.daily_sales over silver.orders with a constraint that amount is never negative, refreshed daily.

Change the item names to your own. Claude asks before it creates or changes anything.

e2e-medallion-architecture

Designs and builds a full medallion platform — workspace and lakehouse layout, ingestion into Bronze, PySpark + Delta transforms for Silver and Gold, per-layer Spark tuning, MLV vs notebook choices, and the Direct Lake handoff to Power BI.

Comes with: spark-cli, semantic-model-authoring. This skill refers to them, so they're added to your install automatically.

Use cases

  1. Design first

    I have CRM, ERP and web-analytics sources. Propose a medallion architecture in Fabric: workspaces, lakehouses, tables per layer and how to orchestrate it. Don't create anything yet.

  2. Build it

    Build a Bronze/Silver/Gold medallion for the NYC taxi sample in workspace 'Taxi-Dev': schema-enabled lakehouse, one notebook per layer, and run them end to end.

  3. Review an existing setup

    Review the lakehouses and notebooks in workspace 'Sales-Dev' against medallion best practices and list what you'd change.

Change the item names to your own. Claude asks before it creates or changes anything.

Dataflows Gen2, Power Query M, and the variables your pipelines consume.

dataflows-cli

Creates and edits Dataflow Gen2 items — Power Query M, connections, output destinations — previews queries before saving, reads refresh history, and upgrades Gen1 dataflows to Gen2.

Use cases

  1. Create a dataflow

    Create a Dataflow Gen2 in 'Sales-Dev' that loads the Customers table from our Azure SQL database into lh_bronze, and show me a preview before saving.

  2. Why did it fail?

    Show the refresh history of dataflow df_customers and explain the last failure.

  3. Gen1 → Gen2 upgrade

    Assess whether the Gen1 dataflows in workspace 'Finance' are safe to upgrade to Gen2, then upgrade the ones with no risks.

Change the item names to your own. Claude asks before it creates or changes anything.

variable-library-cli

Manages Variable Libraries and their value sets (dev / test / prod), and knows the exact syntax to reference a variable from pipelines, notebooks (notebookutils), Dataflow Gen2, Copy Jobs and shortcuts.

Use cases

  1. Remove hard-coded values

    Create a Variable Library 'vl_config' with a storage_account and a sql_server variable, with dev and prod value sets.

  2. Use it in a notebook

    Show me how to read storage_account from vl_config in a PySpark notebook, and update nb_ingest to use it.

  3. Switch environments

    Which value set is active in workspace 'Sales-Test'? Switch it to 'test'.

Change the item names to your own. Claude asks before it creates or changes anything.

T-SQL on Warehouses, SQL endpoints, mirrored DBs and Fabric SQL database.

sqldw-cli

Works with Warehouses, Lakehouse SQL analytics endpoints and Mirrored Databases — tables, COPY INTO, MERGE, read-only queries, CSV export, and slow or failed query investigation through Query Insights.

Also needs: sqlcmd, which gets added to your command.

Without MCP: Without MCP it runs T-SQL through sqlcmd (the skill's built-in fallback). The Capacity Metrics CU-spike correlation workflow needs the SQL endpoint MCP server, so it won't be available.

Use cases

  1. Create and load a table

    In warehouse wh_sales create a dim_customer table and load it with COPY INTO from the parquet files in lh_bronze/Files/customers.

  2. Query the SQL endpoint

    Using the SQL endpoint of lh_gold, show the top 10 products by revenue this year and export the result to CSV.

  3. Slow queries

    Which queries on wh_sales were the slowest or failed in the last 24 hours? Suggest fixes.

Change the item names to your own. Claude asks before it creates or changes anything.

sqldb-cli

Fabric SQL database, the OLTP SQL Server engine — tables, indexes, procedures, vector columns and similarity search, dacpac deployments, and Query Store / blocking investigation. Runs T-SQL through sqlcmd.

Also needs: sqlcmd, which gets added to your command.

Use cases

  1. Schema changes

    In SQL database sqldb_app add an orders table with a foreign key to customers and an index on order_date.

  2. Deploy a dacpac

    Deploy ./build/App.dacpac to sqldb_app and report what changed.

  3. Performance

    sqldb_app is slow right now. Check for blocked sessions and regressed plans in Query Store.

Change the item names to your own. Claude asks before it creates or changes anything.

Eventstreams, Eventhouse/KQL, schemas, alerts and observability data.

eventhouse-cli

Everything KQL — tables, mappings, functions, retention and caching policies, materialized views, ingestion, and read-only time-series queries and ingestion health checks.

Use cases

  1. Create a table

    In KQL database kql_iot create a telemetry table (device_id, ts, temperature, humidity) with 30-day retention and a JSON ingestion mapping.

  2. Query

    Show average temperature per device in 5-minute bins over the last 6 hours in kql_iot.

  3. Materialized view

    Create a materialized view with the latest reading per device in kql_iot.

Change the item names to your own. Claude asks before it creates or changes anything.

eventstream-cli

Eventstream items end to end — sources, filter and aggregate operators, destinations, routing, retention and throughput — plus read-only checks of an existing topology.

Use cases

  1. Build a stream

    Create an Eventstream es_iot that reads from our Event Hub, filters out devices starting with 'test-' and writes into the telemetry table of kql_iot.

  2. Inspect

    Show the topology of es_iot and check whether every node is healthy.

Change the item names to your own. Claude asks before it creates or changes anything.

eventschemaset-cli

Governs Event Schema Sets — the Fabric registry of event types and their payload schemas. Create, update and delete them, or decode an existing definition.

Use cases

  1. Register a schema

    Create an Event Schema Set 'iot-events' with a TelemetryReading event type (device_id string, ts datetime, temperature double).

  2. Inspect

    List all event types in 'iot-events' and decode their schemas.

Change the item names to your own. Claude asks before it creates or changes anything.

activator-cli

Activator (Reflex) alerts — rules, data sources, conditions and the Teams or email action they trigger — plus decoding of existing alert definitions.

Use cases

  1. Threshold alert

    Alert the 'Ops' Teams channel whenever any device in kql_iot reports a temperature above 90.

  2. Audit alerts

    What Activator alerts exist in workspace 'IoT-Prod' and what does each one do?

Change the item names to your own. Claude asks before it creates or changes anything.

azmon-mirroredcatalogs-operations-cli

Brings Azure Monitor, Application Insights and Log Analytics telemetry into Fabric as Eventhouse external delta tables, correlates it with business data, and builds Real-Time dashboards over it.

Use cases

  1. Onboard telemetry

    Bring the Application Insights data for our checkout API into Fabric so I can query it next to sales.

  2. Business impact

    Did the API latency spike last Tuesday affect orders? Correlate App Insights with gold.orders.

Change the item names to your own. Claude asks before it creates or changes anything.

Semantic models, reports and ontologies — via PBIP files and REST, no MCP.

semantic-model-authoring

Semantic models — tables, measures, relationships, field parameters, DAX, Import / DirectQuery / Direct Lake, refresh, permissions and deployment to Fabric.

Also needs: PBIP project (In Power BI Desktop use File → Save as → Power BI project (.pbip), then run Claude Code in that folder.).

Without MCP: Without MCP it edits the model's TMDL files — in a PBIP folder, or a model it downloads from your workspace — and deploys them back. It can't edit a model live in Power BI Desktop, so save your report as .pbip first (File → Save as → Power BI project).

Use cases

  1. Add measures

    In ./SalesReport.SemanticModel add a Date table marked as date table, and measures for Total Sales, Sales YoY and Sales YoY %.

  2. Best-practice review

    Review the semantic model in ./SalesReport.SemanticModel against best practices and fix the high-severity findings.

  3. Deploy and refresh

    Deploy ./SalesReport.SemanticModel to workspace 'Sales-Dev' and trigger a refresh.

Change the item names to your own. Claude asks before it creates or changes anything.

powerbi-report-cli

Plans and designs reports — requirements, page archetypes, chart choice, layout — edits local PBIR/PBIP pages and visuals, validates them, and publishes or rebinds report items in Fabric.

Also needs: PBIP project (In Power BI Desktop use File → Save as → Power BI project (.pbip), then run Claude Code in that folder.).

Use cases

  1. Plan a report

    Plan an executive sales report over the SalesReport semantic model: who it's for, pages and visuals. Wait for my approval before building.

  2. Edit a page

    In ./SalesReport.Report add a page 'Overview' with 4 KPI cards, a monthly sales trend and a slicer for region.

  3. Publish

    Publish ./SalesReport to workspace 'Sales-Dev' and bind it to the semantic model there.

Change the item names to your own. Claude asks before it creates or changes anything.

fabriciq-ontology-cli

Fabric IQ Ontology items — entity and relationship types, bindings to your lakehouse tables, and exploration of schema, lineage and graph walks.

Use cases

  1. Model the business

    Create an ontology with Customer, Order and Product entities bound to the gold tables in lh_gold, with the relationships between them.

  2. Explore

    Summarize the ontology 'sales-ontology' and show how Customer connects to Product.

Change the item names to your own. Claude asks before it creates or changes anything.

Git sync, stage promotion, catalog hygiene and finding things.

git-integration-operations-cli

The Fabric Git integration lifecycle — connect a workspace to Azure DevOps or GitHub, commit, update from Git, check sync status, resolve conflicts, and automate sync with a service principal.

Also needs: Fabric CLI (optional), which gets added to your command.

Use cases

  1. Connect to Git

    Connect workspace 'Sales-Dev' to the main branch of our GitHub repo contoso/fabric-sales, folder /fabric.

  2. Commit

    Commit all pending changes in 'Sales-Dev' with the message 'Add silver notebooks'.

  3. Fix a conflict

    'Sales-Dev' says it has conflicts with Git. Show me what conflicts and help me resolve them.

Change the item names to your own. Claude asks before it creates or changes anything.

deployment-pipelines-authoring-cli

Deployment pipelines for ALM — create stages, assign workspaces, deploy everything or only selected or changed items, poll the deployment, and manage stage permissions.

Use cases

  1. Create a pipeline

    Create a deployment pipeline 'Sales' with Dev, Test and Prod stages assigned to the Sales-Dev, Sales-Test and Sales-Prod workspaces.

  2. Selective deploy

    Deploy only the items that changed from Dev to Test in the 'Sales' pipeline and wait until it finishes.

Change the item names to your own. Claude asks before it creates or changes anything.

onelake-catalog-govern-cli

Audits OneLake catalog health through the Fabric Admin and Power BI APIs — domains, workspace assignment, capacities, sensitivity labels, tags, descriptions, ownership — and applies guarded fixes.

Use cases

  1. Audit

    Find every item in the Finance domain with no description, no sensitivity label or no endorsement.

  2. Remediate

    Assign the workspaces whose names start with 'FIN-' to the Finance domain. Show me the plan before changing anything.

Change the item names to your own. Claude asks before it creates or changes anything.

search-consumption-cli

Finds Fabric items across every workspace by name or type with the Catalog Search API, and returns the workspace and item IDs other skills need.

Use cases

  1. Where is it?

    Where is the lakehouse called lh_sales_gold?

  2. Inventory

    List every semantic model in the tenant with its workspace, and save it to models.csv.

Change the item names to your own. Claude asks before it creates or changes anything.

Move off Databricks, Synapse or HDInsight — and size the capacity first.

databricks-migration

Ports Databricks notebooks and jobs — dbutils → notebookutils, secret scopes → Key Vault, DBFS → OneLake, Unity Catalog → schema-enabled lakehouses, Jobs and DLT → Spark Job Definitions and pipelines.

Comes with: spark-cli, sqldw-cli. This skill refers to them, so they're added to your install automatically.

Use cases

  1. Assess

    Assess the notebooks in ./databricks-export and list everything that needs changing to run in Fabric, by severity.

  2. Convert

    Convert ./databricks-export/etl_orders.py to a Fabric notebook and upload it to workspace 'Migration-Dev'.

Change the item names to your own. Claude asks before it creates or changes anything.

synapse-migration

Ports Synapse Spark pools and notebooks, Lake Databases, Linked Services, and Dedicated SQL Pool schema and stored procedures to Lakehouse Spark SQL — from a live workspace or an offline DACPAC.

Use cases

  1. Assess

    Assess Synapse workspace syn-prod for migration to Fabric: what maps directly and what needs rework?

  2. Migrate procedures

    Migrate the stored procedures in ./dw.dacpac to Spark SQL in lakehouse lh_dw.

Change the item names to your own. Claude asks before it creates or changes anything.

pipeline-migration

Migrates Synapse Data Factory pipelines to Fabric Data Factory — linked services → connections, datasets inlined into activities, global parameters → Variable Libraries, Synapse notebooks → Fabric notebooks.

Use cases

  1. Assess

    Assess the pipelines in Synapse workspace syn-prod and tell me which activities won't migrate cleanly.

  2. Migrate

    Migrate pipeline pl_daily_load from syn-prod to workspace 'Migration-Dev' in Fabric.

Change the item names to your own. Claude asks before it creates or changes anything.

hdinsight-migration

Ports HDInsight Spark and Hive — HiveContext → SparkSession, WASB/ABFS paths → OneLake shortcuts, Hive DDL → Delta tables, Oozie workflows → pipeline activities and triggers.

Comes with: spark-cli, sqldw-cli. This skill refers to them, so they're added to your install automatically.

Use cases

  1. Convert Spark code

    Convert ./hdi/jobs/aggregate.py from HDInsight to a Fabric notebook.

  2. Oozie workflows

    Translate ./hdi/oozie/workflow.xml into a Fabric pipeline design.

Change the item names to your own. Claude asks before it creates or changes anything.

e2e-fabric-cost-estimation

Profiles your current Spark, SQL, Power BI and real-time workloads, maps them to Fabric capacity units, recommends an F-SKU and compares Reserved, Pay-As-You-Go and Autoscale pricing.

Use cases

  1. Size a capacity

    We run 40 Databricks jobs a day on 8-node clusters plus 30 Power BI reports. What Fabric capacity do we need and what will it cost?

  2. Compare billing

    Compare reserved vs pay-as-you-go for an F64 running 12 hours a day in West Europe.

Change the item names to your own. Claude asks before it creates or changes anything.

2

Copy your install command

Paste it into a terminal and run it. It installs the tools, downloads the skills (or updates them if you already have them), copies only the skills you picked, and signs you in to Azure. You'll need Claude Code and git.

No skills selected yet. Go back to step 1 and press + Add on the skills you want.

3

Start Claude Code and ask

Open a terminal in your project folder, start Claude Code, and paste one of the use-case prompts. To confirm the skills loaded, run /skills or ask "which Fabric skills do you have?"

claude

If Claude Code was already open when you installed, restart it so it picks up the new skills.

More options

Update or add more skills later

Pick the skills again, including the ones you already have, and re-run the command. It pulls the latest version and replaces the old folders, so running it again is safe.

Uninstall a skill

Delete its folder. Nothing else was registered.

Remove-Item "$HOME\.claude\skills\spark-cli" -Recurse -Force
rm -rf ~/.claude/skills/spark-cli
Why not the plugin marketplace?

Microsoft's official Claude Code install (/plugin install fabric-skills@fabric-collection) installs every skill and also registers three MCP servers: FabricIQ, Power BI modeling and SQL endpoint. Copying the files gives you only the skills you choose, with no servers. What you give up: fabriciq, which answers natural-language questions over Power BI models, live editing of models open in Power BI Desktop, and the Capacity Metrics correlation in sqldw-cli.

Sources