Labsco
microsoft logo

fabric-cli-core

โœ“ Officialโ˜… 150

by microsoft ยท part of microsoft/fabric-cli

Use Microsoft Fabric CLI (fab) to manage workspaces, semantic models, reports, notebooks, and Fabric resources. Activate when users mention fab, Fabric CLI, or need to automate Fabric operations.

๐Ÿ”ฅ๐Ÿ”ฅ๐Ÿ”ฅโœ“ VerifiedAccount requiredNeeds API keys
๐Ÿงฐ Not standalone. This skill ships with microsoft/fabric-cli and only works together with that tool โ€” install the tool first, then add this skill.

This is the playbook your agent receives when the skill activates โ€” you don't need to read it to use the skill, but it's here to audit before installing.

Fabric CLI Core

This skill defines safe, consistent defaults for an AI agent helping users operate Microsoft Fabric via the Fabric CLI (fab).

1 - Fabric CLI mental model (paths and entities)

Automation Scripts

Ready-to-use Python scripts for core CLI tasks. Run any script with --help for full options.

ScriptPurposeUsage
health_check.pyVerify CLI installation, auth status, and connectivitypython scripts/health_check.py [--workspace WS]

Scripts are located in the scripts/ folder of this skill.

Paths and Entities

  • Treat Fabric as a filesystem-like hierarchy with consistent dot (.) entity suffixes in paths (e.g., .Workspace, .Folder, .SemanticModel).
  • The hierarchy structure is:
    • Tenant: The top-level container for everything.
    • Workspace: Personal or team workspace holding folders, items, and workspace-level elements.
    • Folder: Container for organizing items within a workspace (supports ~10 levels of nesting).
    • Item: Individual resource within a workspace or folder (e.g., Notebook, SemanticModel, Lakehouse).
    • OneLakeItem: OneLake storage item residing within a Lakehouse (tables, files, etc.).
  • Prefer and generate paths like:
    • /Workspace1.Workspace/Notebook1.Notebook
    • /Workspace1.Workspace/FolderA.Folder/SemanticModel1.SemanticModel
    • /Workspace1.Workspace/FolderA.Folder/lh1.Lakehouse/Tables (OneLakeItem)
  • When a user provides an ambiguous identifier, ask for the full path (or infer with stated assumptions).

2 - Modes (interactive vs command line)

  • Be explicit about which mode a user is in:
    • Interactive mode behaves like a REPL and runs commands without the fab prefix.
    • Command line mode runs one command per invocation and is best for scripts/automation.
  • The selected mode is preserved between sessions. If a user exits and logs back in, the CLI resumes in the same mode last used.
  • When you provide instructions, show commands in command line mode unless the user says they're in interactive mode.

3 - Authentication (public-safe guidance)

  • Prefer these auth patterns and do not invent new flows:
    1. Interactive user: fab auth login (browser/WAM where supported).
    2. Service principal (secret/cert): use environment variables / secure mechanisms; avoid embedding secrets in files.
    3. Service principal (federated credential): use the federated token environment variable (FAB_SPN_FEDERATED_TOKEN) and do not persist the raw token.
    4. Managed identity: supported for Azure-hosted workloads; no credentials required.
  • Never ask users to paste secrets into chat or print them back.

4 - Sensitive data handling (strict)

  • Never log or output tokens, passwords, client secrets, or raw federated tokens.
  • Validate all user inputs that could affect security:
    • Paths: Sanitize file paths and API parameters.
    • GUIDs: Validate resource identifiers before use.
    • JSON: Validate JSON inputs for proper format.
  • If a user shares sensitive strings, advise rotating/regenerating them and moving to secure storage.

5 - Hidden entities and discovery

  • Hidden entities are special resources not normally visible, following a dot-prefixed naming convention (similar to UNIX hidden files).
  • Tenant-level hidden entities (accessed from root):
    • .capacities โ€” fab ls .capacities / fab get .capacities/<name>.Capacity
    • .gateways โ€” fab ls .gateways / fab get .gateways/<name>.Gateway
    • .connections โ€” fab ls .connections / fab get .connections/<name>.Connection
    • .domains โ€” fab ls .domains / fab get .domains/<name>.Domain
  • Workspace-level hidden entities (accessed within a workspace):
    • .managedidentities โ€” fab ls ws1.Workspace/.managedidentities
    • .managedprivateendpoints โ€” fab ls ws1.Workspace/.managedprivateendpoints
    • .externaldatashares โ€” fab ls ws1.Workspace/.externaldatashares
    • .sparkpools โ€” fab ls ws1.Workspace/.sparkpools
  • To show hidden resources, recommend ls -a / ls --all.

7 - Output conventions for the agent

  • Default to concise, runnable steps.
  • When recommending commands, include:
    • Preconditions (auth, correct workspace/path)
    • Expected result
    • How to verify (e.g., follow-up fab ls / fab get)

8 - Safety defaults

  • Ask before suggesting commands that delete, overwrite, or change access/permissions.
  • If the user explicitly confirms, proceed with a clear rollback note when possible.

10 - Critical operational rules

  • First run: Always run fab auth status to verify authentication before executing commands. If not authenticated, ask the user to run fab auth login.
  • Learn before executing: Always use fab --help and fab <command> --help the first time you use a command to understand its syntax.
  • Start simple: Try the basic fab command alone first before piping or chaining.
  • Non-interactive mode: Use fab in command-line mode when working with coding agents. Interactive mode doesn't work with automation.
  • Force flag: Use -f when executing commands if the flag is available to run non-interactively (skips confirmation prompts).
  • Verify before acting: If workspace or item name is unclear, ask the user first, then verify with fab ls or fab exists before proceeding.
  • Permission errors: If a command is blocked by permissions, stop and ask the user for clarification; never try to circumvent it.

11 - Common item types

ExtensionDescription
.WorkspaceWorkspace container
.FolderFolder within workspace
.SemanticModelPower BI dataset/semantic model
.ReportPower BI report
.DashboardPower BI dashboard
.NotebookFabric notebook
.LakehouseLakehouse
.WarehouseData warehouse
.DataPipelineData pipeline
.SparkJobDefinitionSpark job definition
.EventstreamReal-time event stream
.KQLDatabaseKQL database
.MLModelML model
.MLExperimentML experiment
.CapacityFabric capacity (hidden)
.GatewayData gateway (hidden)
.ConnectionConnection (hidden)

Use fab desc .<ItemType> to explore any item type.

12 - Command references

For detailed command syntax and working examples, see: