REPO PACKAGE
microsoft skills
OFFICIAL186skills
9groups
16ready to use
LABSCO SUMMARY
Five of the nine sets are per-language Azure SDK guides, each teaching an agent the current shape of one SDK — the client construction, the auth pattern, the retry and pagination idioms — which is what you want when you are writing code against Azure and tired of the model guessing at an API that changed last quarter.
The other four sets are different in kind. azure-skills (32) is operational: cost queries, diagnostics, deployment gates, AKS sizing, capacity hunting. deep-wiki (10) documents a repository you already have — the only set here that does useful work without an Azure account. microsoft-365-agents-toolkit (6) builds Teams and Copilot agents through the ATK CLI. And a loose skills set (11) holds the odds and ends: a KQL reference, an MCP server authoring guide, a Sysinternals debug capture tool.
READ THE FULL ANALYSIS
Where azure-skills comes from. Those 32 are a copy rather than original
work in this repository. Microsoft's own note says the plugin at
.github/plugins/azure-skills/ is copied from microsoft/github-copilot-for-azure,
and that changes to it should be made in that upstream repository rather than
here — so a fix you want in one of these 32 has to land there first.
Who this is for. Engineers already building on Azure, who want their agent to stop hallucinating SDK calls. If you are not on Azure, roughly ten of these 187 will be useful to you and the rest will not.
What it costs you to start. We read all 187 and recorded what each one needs before it works — the labels are on every skill below and you can filter by them. The short version: the large majority need an Azure subscription and credentials, because the thing they produce is only usable against a live Azure resource. A small number are pure guidance that runs on nothing. That distinction is the reason the labels exist: "install it and go" is rare here, and the pages that claim it have been checked one at a time.
Freshness. Microsoft moves this repository quickly — 90 of the 187 skill
documents changed between our July and September captures, and three were
deleted outright. We re-read the whole package in September 2026. Several
skills carry Microsoft's own deprecation notices (the Java Monitor exporters,
azure-communication-callingserver, and the retired Playwright Testing
package); where they do, the skill says so and we have left that in view
rather than tidying it away.
One defect worth knowing. slack-to-teams advertises "100+ micro-expert
files" and routes to ten docs/*.md companions that are not in the
repository. We checked the full git tree; only its SKILL.md is there. That
is Microsoft's omission, not a gap in what we captured.
ALSO IN THIS PACKAGE
AGENT PERSONALITIES
6 itemsAGENT PERSONALITY
Backend Developer
FastAPI and Pydantic house style: a five-class model ladder, two auth dependencies, and a service layer that checks whether Cosmos is reachable before it uses it. Written for a Microsoft demo app called CoreAI DIY rather than as a general Azure role, and its examples use the Pydantic v1 syntax it says it has left behind.
AGENT PERSONALITY
Frontend Developer
React Flow, Zustand and Tailwind conventions, down to the seven node types of the CoreAI DIY canvas. Not a word of Azure in it, and the design tokens it insists on are defined nowhere in the repository.
AGENT PERSONALITY
Infrastructure Engineer
The genuinely Azure one: Bicep modules, azd up and down, reading Container Apps logs. Still written for CoreAI DIY — it hardcodes that app's database name — and its Bicep sample sets environmentId where the property is managedEnvironmentId.
AGENT PERSONALITY
Planner
Explores the codebase and writes a plan under six fixed headings, changing nothing, then hands off to the backend, frontend or infrastructure persona. Its six reference paths, the PRD included, do not exist in this repository.
AGENT PERSONALITY
Presenter Mode Developer
A feature spec for one screen of the CoreAI DIY canvas: viewing versus editing mode, arrow-key navigation, a teleprompter. No Azure, and the components it documents are not in the repository.
AGENT PERSONALITY
Project Scaffolder
The one persona written for anyone: it asks for a project name and scaffolds a React, FastAPI and azd application on Azure Container Apps. It loads three of this package's own skills, one of which is not where it looks for it.
PROMPT TEMPLATES
5 itemsPROMPT TEMPLATE
Add API Endpoint
Writes a REST endpoint across six files, from the Pydantic models to the TypeScript client. Three of the functions it generates are left as pass, and its placeholders are Python expressions inside ${} that no templating engine substitutes.
PROMPT TEMPLATE
Code Review Checklist
A sixty-item review checklist for React and FastAPI work, ending in three commands to run. No Azure, and it cites an AGENTS.md that this repository actually spells Agents.md.
PROMPT TEMPLATE
Create New Node Type
Adds a React Flow node type and wires it into the store, the barrel file and the canvas. No Azure, and the component it emits carries two unused imports, which fails the lint gate this same package sets.
PROMPT TEMPLATE
Create Zustand Store
A Zustand store with the middleware, the loading and error fields and a reset action. No Azure, and the selector it demonstrates returns a fresh object on every render — the re-render it says it prevents.
PROMPT TEMPLATE
Scaffold Foundry App
The big one: a thousand lines that emit about twenty-five files for an Azure AI Foundry application, Bicep and CI included. Genuinely Azure, but it references three files it never writes and its own last code fence is never closed.
PREPARED MCP SERVERS
14 itemsMCP SERVER
aspire
stdio · aspire mcp startMCP SERVER
chrome-devtools
Official Chrome DevTools MCP server for controlling and inspecting a live Chrome browser from coding agents such as Gemini, Claude, Cursor, and Copilot.
stdio · npx chrome-devtools-mcp@latestMCP SERVER
context7
Provides up-to-date, version-specific documentation and code examples for libraries directly into your prompt.
stdio · npx -y @upstash/context7-mcp@latestMCP SERVER
deepwiki
Three tools that answer questions about any public GitHub repository, free and with no account at all.
http · https://mcp.deepwiki.com/sseMCP SERVER
ESLint
stdio · npx eslint --mcpMCP SERVER
github
Official GitHub MCP server for repository search, issues, pull requests, code context, and GitHub workflows in MCP-compatible AI clients.
http · https://api.githubcopilot.com/mcp/MCP SERVER
huggingface
Four tools over the Hub: search models, datasets and spaces, read repo details, and run filesystem operations on a repo.
http · https://huggingface.co/mcpMCP SERVER
markitdown
stdio · uvx markitdown-mcpMCP SERVER
memory
Knowledge graph-based persistent memory system.
stdio · npx -y @modelcontextprotocol/server-memoryMCP SERVER
microsoft-docs
Official Microsoft Learn MCP Server and CLI tool — powering LLMs and AI agents with real-time, trusted Microsoft docs and code samples.
http · https://learn.microsoft.com/api/mcpMCP SERVER
playwright
Official Playwright MCP server for browser automation, page inspection, screenshots, and web interaction from Claude, Cursor, and other AI agents.
stdio · npx @playwright/mcp@latestMCP SERVER
sequentialthinking
Dynamic and reflective problem-solving through thought sequences.
stdio · npx -y @modelcontextprotocol/server-sequential-thinkingMCP SERVER
svelte
Official Svelte MCP server, provides docs and suggestions on the generated code. Our page covers the npm package; Microsoft wires the project's hosted endpoint.
http · https://mcp.svelte.dev/mcpMCP SERVER
terraform
HashiCorp Terraform MCP server for Infrastructure as Code workflows, including provider and module discovery through the Terraform Registry.
stdio · docker run -i --rm hashicorp/terraform-mcp-serverWHAT'S INSIDE
186 showing · 186 totalReady to use
Nothing else to set up — install it and go.
186 of 186
SETUP
Showing all 186 skills
Azure skills31 skills
airunway-aks-setup
Sets up AI Runway on an Azure Kubernetes cluster you already have — from a bare cluster to a running model.
appinsights-instrumentation
How an app running on Azure gets set up to report what it is doing and what is going wrong — automatically where the platform allows it, by changing the code where it does not.
azure-ai
A map of Microsoft's Azure AI services — search, speech, OpenAI models, document reading — pointing you at the right one for the job and the quickest way to call it.
azure-aigateway
Put one front door in front of every AI model, tool and agent your apps call, so the traffic can be measured and controlled in one place instead of app by app.
azure-cloud-migrate
Moving an app off AWS, Heroku or Google Cloud onto Azure: it first reports what the move involves, then rewrites the code for Azure's equivalent services.
azure-compliance
A health check over everything you run in Azure that finds the settings breaking Microsoft's own rules and explains how to put each one right.
azure-compute
The front door for renting computing power on Microsoft Azure. It works out what the job actually needs before anything is created, then hands the request to whichever specialised workflow does that piece.
azure-deploy
Deploys an Azure application that has already been prepared, and gives you the live URLs when it is up.
azure-diagnostics
Works out why an Azure resource is failing — an app, a container, a virtual machine, a cluster or a messaging namespace — and what to do about it.
azure-enterprise-infra-planner
Turns a description of an enterprise workload into an Azure infrastructure plan and the Bicep or Terraform that deploys it.
azure-kubernetes
Gives you a recommended production-ready setup for a new Azure Kubernetes Service cluster, then creates it for you.
azure-kubernetes-automatic-readiness
Tells you whether an existing Azure Kubernetes cluster, and everything running on it, can move to the fully managed Automatic mode — and writes the config changes for whatever cannot.
azure-kusto
Answers questions about the data sitting in an Azure Data Explorer cluster by writing and running Kusto's query language for you.
azure-messaging
Takes a cryptic Event Hubs or Service Bus error — a lost message lock, a link that detached, a processor that quietly stopped — and names the cause and the fix.
azure-prepare
Analyzes a codebase and produces a deployment plan, infrastructure-as-code, and Dockerfiles for Azure, without deploying anything itself.
azure-quotas
Tells you how much quota a subscription has left in each region, and raises the limit when there is not enough.
azure-reliability
Shows where your Azure Functions and App Service resources would fall over — zone redundancy, storage replication, health probes, region failover — and applies the fixes you approve.
azure-resource-lookup
Gives you the list of Azure resources that match what you asked for, across whole subscriptions or a single resource group.
azure-resource-visualizer
Draws a picture of what you have running in Azure — every service in one group, and the lines showing which ones talk to each other.
azure-storage
Works out where your data belongs in Azure — as plain files, a shared drive, a message queue or a simple table — and how fast, how duplicated and how expensive that storage should be.
azure-upgrade
Moves an Azure workload you already run onto a newer plan, tier or service, assessed for readiness first and migrated in steps that leave the original app standing until you say otherwise.
azure-validate
Tells you whether a planned Azure deployment will actually work — configuration, infrastructure templates, role assignments and managed identity permissions all checked — before anything is deployed.
capacity
Tells you where you could actually deploy an Azure OpenAI model: a ranked table of regions and projects showing both the capacity the platform has and the quota your subscription has left.
customize
Takes you through every choice in an Azure OpenAI deployment — version, deployment type, exact capacity, content filter, upgrade policy — and then creates it, after putting the whole configuration in front of you to approve.
deploy-model
Gets an AI model running in Azure, either on sensible defaults or with every setting chosen by hand, after checking there is actually room for it somewhere.
entra-agent-id
Gives every running AI agent its own Microsoft Entra identity, so each one signs in with its own OAuth tokens, its own permissions and its own line in the audit log.
entra-app-registration
Sets an app up so people can sign in to it with their Microsoft work account — or so a background service can let itself in with nobody present — and writes the matching code.
finetuning
Takes you from a raw training set to a fine-tuned model deployed on Microsoft Foundry, with numbers showing whether it actually beat the base model.
microsoft-foundry
One place to take an AI agent on Microsoft Foundry from nothing to running: the project and model underneath it, then building, deploying, testing and repairing the agent itself.
preset
Gets an Azure OpenAI model deployed into a region that can actually take it, instead of failing on capacity and starting over.
python-appservice-deploy
Takes a Python web app and gets it running on Microsoft's hosting, creating the hosting itself first if the app has nowhere to land yet.
Azure SDK / Python40 skills
agent-framework-azure-ai-py
Build an AI assistant from Python that is created and stored on Microsoft's Azure AI Foundry service, so it stays there between sessions instead of living inside your program.
azure-ai-contentsafety-py
From Python you get a severity score for a piece of text or an image across hate, sexual, violence and self-harm, so the app can allow, flag or block it.
azure-ai-contentunderstanding-py
One service for four kinds of file — documents, images, audio and video — that turns each one into something a program can actually work with.
azure-ai-language-conversations-py
Work out what someone actually wants from the sentence they typed — book a flight, cancel an order — using a model you trained in Azure on your own example phrases.
azure-ai-ml-py
Drives an Azure Machine Learning workspace from Python instead of clicking through the studio.
azure-ai-projects-py
Microsoft Foundry keeps an AI assistant and everything it needs — models, data, test results — together in one project; this is the Python client for that project.
azure-ai-textanalytics-py
Reads a block of ordinary writing from Python and tells you what is in it — who is mentioned, how it feels, what ought to be hidden — using Azure's ready-made language models.
azure-ai-transcription-py
Turns recorded or live speech into text from Python.
azure-ai-translation-document-py
Translates whole documents — Word, PDF, PowerPoint, Excel, HTML and more — into other languages with their formatting left intact.
azure-ai-translation-text-py
Translates text into one or more languages from Python code.
azure-ai-vision-imageanalysis-py
Sends a picture to Azure from Python and gets a written account of what is in it back, so the program can act on the contents.
azure-ai-voicelive-py
A live voice conversation driven from Python: the user talks into the microphone and the model answers out loud, with no text step in between.
azure-appconfiguration-py
Settings and feature switches live in one place in Azure rather than in a file beside the code, and a Python program pulls them from there as it runs.
azure-containerregistry-py
From Python you can see everything held in an Azure Container Registry — repositories, tags and image manifests — and delete what is no longer wanted.
azure-cosmos-db-py
How to wire a Python web API to Azure's Cosmos database and keep all the data code in its own layer, instead of scattering queries through the app.
azure-cosmos-py
Microsoft's Cosmos DB keeps data as free-form documents rather than rows in tables; this is how Python code puts documents in and gets them back out.
azure-data-tables-py
Reads and writes rows in Azure's table service from Python, with a matching async client for code that cannot sit and wait.
azure-eventgrid-py
Python code that posts a message to Azure Event Grid whenever something happens in your app, and lets Azure deliver it to everything that subscribed.
azure-eventhub-py
Python's client for Azure Event Hubs, the service that swallows huge volumes of small messages — written once for ordinary blocking code and once for asyncio.
azure-identity-py
Signs a Python app in to Azure with no password in the code, and can pass the same short-lived token to a library that is not an Azure one, such as OpenAI's.
azure-keyvault-py
All three things an Azure Key Vault holds — passwords, encryption keys and certificates — reached from Python, one client and one package for each.
azure-messaging-webpubsubservice-py
Azure Web PubSub from Python on both sides of a live connection: the server that sends updates out, and a Python client that sits and listens for them.
azure-mgmt-apicenter-py
Registers each of your organisation's APIs into Azure API Center from Python — its versions, its spec file, who owns it, and which environment it is running in.
azure-mgmt-apimanagement-py
Azure API Management from Python: publish an API you already have through Azure's gateway, so outside callers reach it with a key instead of hitting your server directly.
azure-mgmt-botservice-py
Sets up a chatbot's Azure side from Python — where it listens, which apps people can reach it in, and the sign-in that lets it act on a user's behalf elsewhere.
azure-mgmt-fabric-py
Microsoft Fabric needs a pool of compute behind it; this is the Python code that creates that pool, changes its size, and pauses it when nobody is using it.
azure-monitor-ingestion-py
Puts log records a Python program produced into an Azure Monitor table — the same place the logs from Azure's own services land.
azure-monitor-opentelemetry-exporter-py
The hands-on way to get a Python app's timings, counters and log lines into Azure's Application Insights — you assemble the pipeline yourself instead of taking the all-in-one package.
azure-monitor-opentelemetry-py
One setup call, and a Python app starts reporting on itself to Azure — how long each request took, what it called out to, and what it logged — with no instrumentation code written by hand.
azure-monitor-query-py
Pulls Azure's stored logs and performance figures back into Python, either as the rows a query returns or as a single measurement sampled over an interval.
azure-search-documents-py
Everything a Python app needs to search your own files — getting them in, keeping the index fresh, and finding the right passage for a question.
azure-servicebus-py
A Python program drops jobs into a queue and picks them up again later, with each one tracked until it is finished, retried, or set aside as a failure.
azure-speech-to-text-rest-py
Turns a spoken audio file into text with one HTTPS request to Azure Speech, coming back with the transcript and a recognition status you can branch on.
azure-storage-blob-py
Keeps a Python app's files in Azure instead of on its own disk, so any machine running the code can read and write the same data.
azure-storage-file-datalake-py
Gives Python code real folders in cloud storage — nested paths, renaming, and who is allowed to touch each file — instead of one flat pile of everything.
azure-storage-file-share-py
Puts files on a shared drive that lives in the cloud, and works with them from Python without ever connecting that drive to a computer.
azure-storage-queue-py
A waiting list of jobs that one Python program adds to and others work through — nothing gets picked up twice, and nothing is lost if a worker dies halfway.
fastapi-router-py
Adds a fresh set of web-API endpoints for one kind of item — list them, create one, change it, delete it — built the same way as the rest of the project.
m365-agents-py
A Python service that answers people's messages in Teams and Microsoft 365, typing its replies out as the model writes them rather than after a long silence.
pydantic-models-py
Writes out the set of data shapes one part of an API needs — what comes in, what goes back, what gets stored — so they agree with each other instead of drifting apart.
Azure SDK / .NET29 skills
azure-ai-agents-persistent-dotnet
Gives your .NET app an agent that lives on the Azure AI Agent Service and keeps its own conversation threads.
azure-ai-document-intelligence-dotnet
Reads a PDF or a scanned form and hands your .NET code its text, its tables and its named fields, each with a confidence score.
azure-ai-openai-dotnet
Your .NET app reaches every model deployed on an Azure OpenAI resource through one client.
azure-ai-projects-dotnet
An Azure AI Foundry project is the workspace holding your deployed models, your data and the agents built on them — this is the .NET client that reaches all of it.
azure-ai-voicelive-dotnet
Your .NET app holds a spoken back-and-forth conversation with an Azure model in real time, streaming microphone audio up and the spoken reply back.
azure-eventgrid-dotnet
Sends notifications that something happened — an order was placed, a file was uploaded — out of a .NET app through Azure Event Grid, and receives them at the other end.
azure-eventhub-dotnet
Pushes a stream of small messages — sensor readings, clicks, log lines — out of a .NET service into Azure Event Hubs, and reads the stream back on the other side.
azure-identity-dotnet
Gets a .NET app signing in to Azure services with Microsoft Entra ID and no hardcoded secret.
azure-maps-search-dotnet
Puts Azure Maps behind a .NET application so it can answer where-something-is, how-to-get-there and what-the-weather-is questions.
azure-mgmt-apicenter-dotnet
Keeps an organization-wide catalog of your APIs in Azure API Center, from code written in C#.
azure-mgmt-apimanagement-dotnet
Creates the gate your APIs sit behind — who may call them, on what plan, how often — and sets the whole thing up from C#.
azure-mgmt-applicationinsights-dotnet
Where an app's performance data lands in Azure, how long it is kept, and the ping tests calling the site from around the world to check it is up — all created from C#.
azure-mgmt-arizeaiobservabilityeval-dotnet
Arize AI is a third-party service for watching how machine-learning models behave once they are live; this registers it as a billable resource inside your own Azure subscription, from C#.
azure-mgmt-botservice-dotnet
Creates an Azure Bot resource from C# and opens the channels users reach it on, from Teams to Web Chat to Direct Line.
azure-mgmt-fabric-dotnet
The block of compute power a Microsoft Fabric workspace runs on — this creates one from C#, resizes it, and pauses or deletes it when it is no longer wanted.
azure-mgmt-mongodbatlas-dotnet
MongoDB Atlas can be bought through Azure and billed on the same invoice; this creates and looks after the Azure-side record that makes that link, from .NET code.
azure-mgmt-weightsandbiases-dotnet
Deploys a Weights & Biases experiment-tracking instance as an Azure resource.
azure-resource-manager-cosmosdb-dotnet
The Cosmos DB account, databases and containers an app's data will sit in, all created from .NET before a single record is written.
azure-resource-manager-durabletask-dotnet
Creates the Azure service that runs an app's long, multi-step workflows, along with the task hubs inside it, ready for the app to connect to later.
azure-resource-manager-mysql-dotnet
A managed MySQL database on Azure, created from .NET code and then kept in shape from there too: bigger when it is busy, switched off when nobody needs it.
azure-resource-manager-playwright-dotnet
Creates and manages the Microsoft Playwright Testing workspaces that cloud browser tests run against.
azure-resource-manager-postgresql-dotnet
Creates a PostgreSQL database server on Azure from .NET code, then manages it afterward — databases, server settings, backups, replicas, and scaling all included.
azure-resource-manager-redis-dotnet
Sets up an Azure Cache for Redis — the fast in-memory store an app keeps in front of its database — and then looks after its access keys, its firewall and when it gets patched.
azure-resource-manager-sql-dotnet
Creates Azure SQL servers, databases, elastic pools, and firewall rules from .NET code.
azure-search-documents-dotnet
How a C# app searches its own data — matching the words someone typed, the meaning behind them, or both at once.
azure-security-keyvault-keys-dotnet
Keeps a C# app's encryption keys locked inside Azure's vault instead of in your own code, and has the vault do the encrypting and signing so the key never leaves it.
azure-servicebus-dotnet
Lets the separate parts of a C# application hand work to each other through a queue, so a job waits safely until something is ready to pick it up.
m365-agents-dotnet
Builds a chatbot that works inside Teams and Microsoft 365, written in C# and hosted as an ordinary web service, with the sign-in and message plumbing already done.
microsoft-azure-webjobs-extensions-authentication-events-dotnet
Runs your own code in the middle of somebody signing in with their Microsoft account — adding extra facts about them, changing what the sign-up form asks for, or sending the one-time code yourself.
Azure SDK / TypeScript25 skills
applicationinsights-web-ts
Puts monitoring inside the web page itself, so problems people actually hit in their browser show up in Microsoft's Application Insights alongside the server's own records.
azure-ai-contentsafety-ts
Azure's moderation service, called from TypeScript: it rates how harmful a message or a picture is and tells you which of your own banned words it hit.
azure-ai-document-intelligence-ts
Invoices, receipts, IDs and other paperwork go in as a PDF or a scan, and come back to a TypeScript program as ordinary data — the amounts, the dates, the tables.
azure-ai-projects-ts
From TypeScript an Azure AI Foundry project is a single object: the agents you run in it, and the resources those agents draw on, all hang off the same client.
azure-ai-translation-ts
Gives a TypeScript or Node app translated text, and translated copies of whole documents, in the languages you ask for.
azure-ai-voicelive-ts
A spoken assistant that can run either on a Node server or inside the browser itself, powered by an Azure model that listens and replies out loud.
azure-appconfiguration-ts
Nothing in this Node or TypeScript app's configuration has to sit in the repository — the values, and the switches that turn features on, come from an Azure store instead.
azure-cosmos-ts
Gets a TypeScript or JavaScript app creating, querying and bulk-writing documents in Azure Cosmos DB.
azure-eventhub-ts
Node and TypeScript code for Azure Event Hubs, where a device or an app fires telemetry in continuously and a subscriber picks it up as it arrives.
azure-identity-ts
Keeps a Node or TypeScript app signed in to Azure — a Visual Studio Code login while you develop, a browser prompt for a real user, a server identity in production.
azure-keyvault-keys-ts
A Node app's passwords and encryption keys, kept in Azure Key Vault instead of in its config — read back, used for encrypting and signing, rotated, and restored from backup.
azure-keyvault-secrets-ts
Stores and retrieves an application's secrets and cryptographic keys in Azure Key Vault from TypeScript code.
azure-microsoft-playwright-testing-ts
Runs an existing Playwright suite on Microsoft's cloud-hosted browsers instead of local ones.
azure-monitor-opentelemetry-ts
A Node or TypeScript service that reports on itself to Azure: every request it serves and every call it makes out, switched on at startup rather than written into the code that does the work.
azure-postgres-ts
A Node.js or TypeScript app gets a pooled, encrypted connection to Azure Database for PostgreSQL Flexible Server that it can query and run transactions on.
azure-search-documents-ts
Puts a working search box into a TypeScript or Node app, one that finds documents by the words in them and by what they are about.
azure-servicebus-ts
Two parts of a system can hand work to each other without both being up at the same time — Azure Service Bus holds the message in between, and this is the TypeScript side of it.
azure-storage-blob-ts
Cloud storage for a web app's files, driven from TypeScript: how a Node server or a browser page gets a file into Azure and back out again.
azure-storage-file-share-ts
The familiar shared network drive, except it lives in Azure: from TypeScript you can set one up, organise it into folders, and move files on and off it.
azure-storage-queue-ts
Keeps a TypeScript service working through a backlog of queued jobs, including the ones that run long and the ones that will never succeed.
azure-web-pubsub-ts
Live updates in the browser — chat messages, notifications, anything that has to arrive the moment it happens — with Azure holding all those open connections instead of your own server.
frontend-ui-dark-ts
A ready-made dark look for a React app: deep backgrounds, purple accents, frosted-glass panels and movement that settles rather than jumps.
m365-agents-ts
Puts an assistant into Teams or Microsoft 365 from Node and TypeScript — or, if you would rather, into a chat window on a page of your own.
react-flow-node-ts
Adds a new kind of box to a drag-and-drop workflow diagram in a React app — the box itself, the dots that wires connect to, and its entry in the menu.
zustand-store-ts
One agreed way to build the Zustand store that holds a React app's shared data — the same structure, the same separation of the data from the functions that change it, every time.
Azure SDK / Java26 skills
azure-ai-agents-persistent-java
A bare-bones way to run an AI assistant on Azure from Java: no framework on top, so every step of the conversation is a call you make yourself.
azure-ai-anomalydetector-java
Finds the points in a time series that do not belong, from Java.
azure-ai-contentsafety-java
Scores text and images for harmful content from Java, so the app can allow or block them on a severity threshold.
azure-ai-formrecognizer-java
Pulls the text, the tables and the filled-in fields out of a document from Java, using the current Document Intelligence package rather than the older Form Recognizer one it replaced.
azure-ai-projects-java
Housekeeping for an Azure AI Foundry workspace from Java — taking stock of what the project is wired up to, and what has been registered inside it.
azure-ai-vision-imageanalysis-java
Lets Java code work out what a picture actually shows, rather than treating it as a file it can only store and pass around.
azure-ai-voicelive-java
Speech in, speech out — a Java program keeps an audio channel open to an Azure model so the user can just talk, and hear the answer spoken back.
azure-appconfiguration-java
A Java application reads its settings and feature flags out of one central App Configuration store instead of out of local files.
azure-communication-callautomation-java
A Java service runs the call flow for real phone calls from your own server code.
azure-communication-callingserver-java
Maps the classes and methods of the retired CallingServer SDK onto their Call Automation equivalents so an old Java codebase can be moved across.
azure-communication-chat-java
A Java app gets chat threads where people send messages, see who has read them, and watch each other type.
azure-communication-common-java
Before a Java app can chat, call or text through Azure Communication Services, it has to say which user it is acting for and prove it — that is what this shared piece supplies.
azure-communication-sms-java
Sends a text message from Java to one number or to a whole list of them.
azure-compute-batch-java
Runs command-line tasks across a pool of Azure VMs from Java.
azure-cosmos-java
Reads and writes JSON documents in a Cosmos DB NoSQL database from Java, through a blocking or a reactive client.
azure-data-tables-java
Stores and looks up rows of data in Azure's table service from Java, where a row is found by two keys instead of by a SQL query.
azure-eventgrid-java
Java's end of Azure Event Grid — announce that something happened so other services can react, and make sense of the announcements arriving from Azure's own services.
azure-eventhub-java
Java on both ends of an Azure Event Hubs stream: the code that writes a firehose of events in, and the code that reads them back out.
azure-identity-java
Every way a Java program can prove who it is to Azure — a developer's own login, a managed identity, a certificate, a browser prompt — and which one belongs where.
azure-messaging-webpubsub-java
A Java backend that can push a chat message, a live score or an alert straight to a browser that already has the page open, through Azure Web PubSub.
azure-monitor-ingestion-java
Uploads a Java application's own log lines into Azure's log store, so they sit alongside the logs Azure already collects for you.
azure-monitor-opentelemetry-exporter-java
Sends the timings, counts and errors a Java app already records with OpenTelemetry on to Application Insights, Azure's monitoring service.
azure-monitor-query-java
Asks Azure for data it has already collected — lines out of a log store, or one measurement tracked over time — and hands the answer back to Java code.
azure-security-keyvault-keys-java
A Java app gets its encryption and signing done by Azure's vault, which holds the keys so they are never copied into the application itself.
azure-security-keyvault-secrets-java
Passwords, API keys, and connection strings live in Key Vault rather than in config files, and a Java app reads them at run time.
azure-storage-blob-java
Stores files in the cloud from a Java app and fetches them back, including the big ones that have to travel a piece at a time.
Azure SDK / Rust8 skills
azure-cosmos-rust
Wires a Rust program up to Azure Cosmos DB, Microsoft's cloud database, using the official package rather than the community look-alikes that share its name.
azure-eventhub-rust
Rust side of Azure Event Hubs: open a sender and push events in singly or in batches, then open a reader on one slice of the stream and walk it event by event.
azure-identity-rust
Rust has no one-size-fits-all Azure sign-in, so this pairs the developer's own login while coding with an identity attached to the app once it is deployed.
azure-keyvault-certificates-rust
Issues and looks after the certificates that prove a server is who it says it is, kept inside Azure Key Vault and handled from Rust.
azure-keyvault-keys-rust
Encryption keys that live inside Azure Key Vault and never come out: Rust code creates them and then asks the vault to do the locking and unlocking on its behalf.
azure-keyvault-secrets-rust
Gets a Rust app storing and reading passwords, API keys and connection strings in Azure Key Vault.
azure-storage-blob-rust
Rust code that puts files into Azure's cloud storage and reads them back, signing in with a work account rather than a stored key.
azure-storage-queue-rust
Lets one Rust program leave a short message for another to pick up later, using Azure's queue as the letterbox between them.
Deep Wiki10 skillsCalled with /deep-wiki:* commands.
wiki-ado-convert
Converts a set of documentation pages so they display properly in an Azure DevOps wiki, diagrams included, writing the converted copies alongside rather than over the originals.
wiki-agents-md
Every folder of your repo that matters ends up with an AGENTS.md — the context file a coding agent reads before it touches anything, written out of this project's own files rather than boilerplate. A short CLAUDE.md is dropped beside each one pointing there.
wiki-architect
Turns a codebase into the table of contents for its wiki: a nested catalogue of sections, each carrying a writing prompt that already cites the real files it should be built from.
wiki-changelog
Turns your repository's commit history into a changelog a reader can actually use — grouped by day and week, and sorted into features, fixes, refactors and docs.
wiki-llms-txt
Packages a project's documentation so an AI can take it all in at once — an index at the top of the project and inside the wiki, plus one file holding every page in full.
wiki-onboarding
A codebase gets an onboarding folder holding four written guides — one each for a new contributor, a staff engineer, an engineering leader and a product manager — and a hub page that sends each reader to theirs.
wiki-page-writer
One component or subsystem comes out fully documented — why it exists and how it really works — with diagrams throughout and a file-and-line citation standing behind every claim.
wiki-qa
Answers a question about a codebase out of the files that actually decide the answer, with each point cited down to the file and line it came from.
wiki-researcher
A deep investigation of one part of a codebase where nothing is claimed unless it was actually read — each finding names the file it came from and says how confident it is.
wiki-vitepress
Takes documentation that exists only as a pile of text files and turns it into a real website people can click around in, dark-themed out of the box.
M365 Agents Toolkit6 skills
declarative-agent-developer
Builds and maintains a custom assistant inside Microsoft 365 Copilot, redeploying it after every edit so there is always a live link to try what just changed.
install-atk
Leaves you with the Microsoft 365 Agents Toolkit ready to use: the command-line tool confirmed working, and the VS Code extension installed if you want it.
m365-agent-evaluator
Puts a deployed Microsoft 365 Copilot agent through a scored test suite and tells you which answers failed and why.
slack-to-teams
Moves a chat bot from Slack over to Microsoft Teams, or keeps one running on both, carrying across the buttons, sign-in and file handling rather than only the messages.
teams-app-developer
Takes a Teams or Microsoft 365 agent from an empty folder to something you can talk to, using Microsoft's Agents Toolkit command-line tool and a sub-skill for each stage.
ui-widget-developer
Puts a small interactive panel — a card, a chart, a form — inside a Microsoft 365 Copilot chat, filled with live data from a service of your own.
Skills11 skills
cloud-solution-architect
An architect's pass over a system before it is built: what overall shape it should take, which Azure service does each job, and why each of those choices was made.
continual-learning
Gives an AI coding assistant a memory that outlives the session it was in, so what it picked up about your project is still there the next time you open it.
copilot-sdk
Instead of talking to GitHub Copilot in an editor, your own program does the talking — it sends the prompts and receives the answers, from Node, Python, Go or .NET.
debugview
Captures debug output from Windows programs and kernel drivers as it happens, from the command line.
frontend-design-review
Hands back a prioritised list of what is wrong with a UI — blocking, major, minor — or, the other way round, a new interface with a deliberate aesthetic instead of the usual AI-generated look.
github-issue-creator
Turns a messy bug report — an error paste, a voice note, a screenshot — into a clean, structured issue saved as a dated markdown file in the repo.
kql
Searching huge tables of logs and events in Microsoft's data tools, with queries that come back correct, quick, and without running out of memory.
mcp-builder
Builds the connector that lets an AI assistant use an outside service — planned first, then written, tested, and scored against questions you check yourself.
microsoft-docs
Answers questions about Microsoft products from the official Microsoft Learn documentation rather than from memory.
podcast-generation
Turns a block of text into spoken, podcast-style audio your app can play back, narrated by Azure OpenAI's realtime voice model.
skill-creator
Writes a new instruction pack for an AI coding assistant, usually about an Azure library — trimmed to size, tested, and filed where this collection expects it.
No skill in this package matches that.
HOW TO GET IT
WHOLE PACKAGEInstall the whole package (186 skills)
npx skills add microsoft/skillsMicrosoft asks you not to take all of them: "Loading all skills causes context rot: diluted attention, wasted tokens, conflated patterns. Only copy skills essential for your current project." The command opens a wizard, so you pick. What you pick lands in your agent's own directory — .github/skills/ for GitHub Copilot — and is symlinked if you run more than one agent.
SINGLE SKILLOr install a single skill
npx skills add microsoft/skills --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.
COPILOT CLI PLUGINOr install it as a Copilot CLI plugin
/plugin marketplace add microsoft/skills/plugin install deep-wiki@skills/plugin install azure@skillsTyped inside Copilot CLI, not in a terminal. deep-wiki then answers thirteen of its own: /deep-wiki:generate writes a whole wiki for the repository you are in, :ask answers a question about it, :onboard writes the onboarding guides — and ten more. The azure plugin has no slash commands; it triggers on what you ask for.
MORE INSTALL OPTIONS
If the package is already on disk somewhere, a second project or a second agent can point at that copy instead of installing another one.
One skill, from a copy you have
ln -s /path/to/skills/.github/skills/<name> /path/to/your-project/.github/skills/<name>A second agent, same repository
ln -s ../.github/skills .opencode/skillsClaude, same repository
ln -s ../.github/skills .claude/skills