Skip to main content
hc
Submit
skillsSource-backedReview first Safety ✓ Privacy ✓

Microsoft Agent Skills

Microsoft-maintained Agent Skills, plugins, custom agents, AGENTS.md templates, and MCP configurations for Azure SDKs, Microsoft Foundry, Microsoft 365 Agents SDK, Copilot SDK, MCP server building, and cloud solution work.

by Microsoft·added 2026-06-18·
Claude CodeCodexWindsurfGeminiCursorCLI
HarnessClaude CodeCodexWindsurfGeminiCursorCLI
Level:expertType:generalVerified:validated
Review first — review before installing

Open the source and read safety notes before installing.

Safety notes

  • Some skills guide Azure or Foundry provisioning, hosted agents, model deployments, toolboxes, workflows, storage, messaging, identity, Key Vault, monitoring, and governance tasks that can affect live cloud resources.
  • Validate resource names, subscriptions, tenants, regions, RBAC, quotas, and costs before approving generated commands or infrastructure changes.
  • The upstream README describes the repository as work in progress with active skill updates and expanding tests.
  • Use skills selectively; loading all skills can dilute context and produce mixed SDK patterns.
  • Skills and MCP configs may invoke external documentation, GitHub, browser automation, Azure CLI, azd, or SDK commands depending on the selected workflow.

Privacy notes

  • Azure subscription IDs, tenant IDs, resource names, logs, traces, SDK error output, repository code, architecture notes, and Foundry project details can enter prompts or tool output.
  • Do not paste service principals, access tokens, connection strings, Key Vault secrets, managed identity credentials, or customer data into public issues, screenshots, or committed configs.
  • Microsoft Docs MCP, Context7, GitHub, browser, or other MCP integrations may send query text and project context to external services depending on local configuration.

Prerequisites

  • Node.js and npx for the skills installer.
  • A coding agent or editor setup that can load Agent Skills from `.github/skills`, `.claude/skills`, `.opencode/skills`, or a supported skills directory.
  • Relevant Azure, Microsoft Foundry, Microsoft 365, GitHub, or SDK credentials for the specific skill being used.
  • Current Microsoft Learn, SDK, and package-version checks before implementing against Azure or Foundry APIs.
  • A narrow skill-selection plan; the upstream README warns against loading every skill at once.

Schema details

Install type
package
Reading time
6 min
Difficulty score
78
Troubleshooting
Yes
Breaking changes
No
Source repository stats
Scope
Source repo
Package metadata
Skill and platform metadata
Skill type
general
Skill level
expert
Verification
validated
Verified at
2026-06-18
Retrieval sources
https://github.com/microsoft/skillshttps://raw.githubusercontent.com/microsoft/skills/main/README.mdhttps://raw.githubusercontent.com/microsoft/skills/main/Agents.mdhttps://microsoft.github.io/skills/https://skills.sh/microsoft/skills
Tested platforms
ClaudeCodexGitHub CopilotOpenCodeCursor
PlatformSupportInstall path
claude-codeNative.claude/skills/<skill-name>/SKILL.md
codexNative.agents/skills/<skill-name>/SKILL.md
windsurfNative.windsurf/skills/<skill-name>/SKILL.md
geminiNative.gemini/skills/<skill-name>/SKILL.md or .agents/skills/<skill-name>/SKILL.md
cursorAdapter.cursor/rules/<skill-name>.mdc
cliManualAGENTS.md or tool-specific context file
Tool listing metadata
Full copyable content
# Install interactively
npx skills add microsoft/skills

# Manual install example
git clone https://github.com/microsoft/skills.git
cp -r agent-skills/.github/skills/azure-cosmos-db-py your-project/.github/skills/

About this resource

Microsoft Agent Skills

microsoft/skills is a Microsoft-maintained Agent Skills repository for coding agents working with Azure SDKs, Microsoft Foundry, Microsoft 365 agents, Copilot SDK patterns, MCP servers, custom agents, and project-level agent instructions.

The repository includes skills, plugins, custom agents, AGENTS.md templates, and MCP configurations. It is useful when an agent needs current Microsoft domain context that is too specific or too fast-moving for model memory alone.

What Is Inside

The upstream README currently advertises a large skill catalog, plus plugins, custom agents, AGENTS.md templates, MCP configs, and documentation.

Area Examples
Core mcp-builder, skill-creator, copilot-sdk, cloud solution architecture
Foundry projects, models, hosted agents, toolboxes, workflows, IQ knowledge bases, managed skills, memory, observability, governance
Python Azure AI Projects, Cosmos DB, AI Search, Storage, Service Bus, Monitor, Key Vault, FastAPI patterns
.NET Azure AI, Search, Key Vault, Resource Manager, Foundry, and service SDK patterns
TypeScript Azure SDK, Aspire, storage, AI, and app patterns
Java Azure SDK and service patterns
Rust Azure and infrastructure-oriented SDK patterns

Installation

Install interactively with the skills CLI:

npx skills add microsoft/skills

The installer lets you select the skills needed for the current project. The README says selected skills are installed into the chosen agent directory and can be symlinked across multiple agent configs.

Manual install is also documented upstream:

git clone https://github.com/microsoft/skills.git
cp -r agent-skills/.github/skills/azure-cosmos-db-py your-project/.github/skills/

Selection Guidance

Do not load the whole catalog by default. The upstream README explicitly warns that loading all skills causes context rot: diluted attention, wasted tokens, and conflated patterns. Select only the Azure SDK, Foundry, Copilot, MCP, or language-specific skills needed for the task.

Before implementation, use fresh sources:

  • Search Microsoft Learn or Microsoft Docs MCP for current API signatures.
  • Verify installed SDK package versions.
  • Check the target language package docs and examples.
  • Use Context7 or other current documentation sources when the selected skill asks for it.

Use Cases

  • Add Azure SDK implementation patterns to a coding agent session.
  • Build or review Microsoft Foundry hosted agents, toolboxes, workflows, memory, observability, or governance flows.
  • Create an MCP server in Python, Node/TypeScript, or .NET with the mcp-builder skill.
  • Use skill-creator to author new project skills with tests and acceptance criteria.
  • Ground GitHub Copilot SDK or Microsoft 365 Agents SDK work in current patterns.
  • Add project-level AGENTS.md guidance for repository-specific agent behavior.

Safety and Privacy

Many Microsoft skills are operational. Depending on the selected skill, an agent may propose Azure CLI, azd, SDK, infrastructure, model deployment, identity, storage, monitoring, or governance changes. Treat those as production changes until proven otherwise.

Review cloud subscription, tenant, region, resource group, RBAC, quota, and cost impact before approving generated commands. Keep credentials in local secret stores or environment variables, not in prompts or committed files.

Troubleshooting

The agent mixes SDK versions

Stop and verify package versions, then re-open the relevant Microsoft Learn or SDK docs. The upstream Agents.md warns that Azure SDKs and Foundry APIs change constantly and stale patterns break code.

The agent loads too much context

Remove unrelated skills from the active project directory. Reinstall only the language, service, or Foundry skills required for the current change.

A skill suggests a risky Azure command

Ask the agent to produce a dry-run plan first, including subscription, tenant, region, resource group, permissions, cost impact, rollback path, and validation command.

Source Review

Verified on 2026-06-18:

  • The upstream README describes Microsoft Agent Skills as skills, custom agents, AGENTS.md templates, and MCP configurations for Azure SDKs and Microsoft AI Foundry.
  • The README documents npx skills add microsoft/skills as the quick-start installer.
  • The README warns to use skills selectively rather than loading all skills.
  • Agents.md says Azure SDKs and Foundry APIs change constantly and instructs agents to search current docs and verify SDK versions before implementing.
  • The repository is MIT licensed and published by microsoft/skills.

Duplicate Check

No existing microsoft/skills, Microsoft Agent Skills, or Microsoft Foundry skills catalog entry was found in content/skills.

#microsoft#azure#foundry#agent-skills#mcp

Source citations

Add this badge to your README

Show that Microsoft Agent Skills is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.

Listed on HeyClaude
[![Listed on HeyClaude](https://heyclau.de/badge/skills/microsoft-agent-skills.svg)](https://heyclau.de/entry/skills/microsoft-agent-skills)

How it compares

Microsoft Agent Skills side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.

FieldMicrosoft Agent Skills

Microsoft-maintained Agent Skills, plugins, custom agents, AGENTS.md templates, and MCP configurations for Azure SDKs, Microsoft Foundry, Microsoft 365 Agents SDK, Copilot SDK, MCP server building, and cloud solution work.

Open dossier 		Microsoft Fabric Skills

Microsoft-maintained Fabric skill bundles for AI coding assistants working with Warehouses, Lakehouses, Spark, Power BI semantic models, Eventhouse/KQL, Eventstreams, Dataflows Gen2, migrations, and medallion architectures.

Open dossier 		Wrangler Deployment Operations Capability Pack Skill

Expert Wrangler CLI deployment skill for release planning, config review, environment checks, version traffic splits, trigger application, and rollback.

Open dossier 		.NET Agent Skills

Microsoft .NET team skill marketplace for AI coding agents working on .NET, C#, ASP.NET Core, Blazor, MAUI, diagnostics, MSBuild, NuGet, upgrades, tests, AI workflows, RAG pipelines, and C# MCP servers.

Open dossier
Trust
Install riskReview firstReview firstReview firstReview first
Notes Safety ✓ Privacy ✓ Safety ✓ Privacy ✓ Safety ✓ Privacy ✓ Safety ✓ Privacy ✓
Categoryskillsskillsskillsskills
Sourcesource-backedsource-backedsource-backedsource-backed
AuthorMicrosoftMicrosoftoktofeesh1.NET Team at Microsoft
Added2026-06-182026-06-182026-06-032026-06-18
Platforms
Claude CodeCodexWindsurfGeminiCursorCLI
Claude CodeCodexWindsurfGeminiCursorCLI
Claude CodeCodexWindsurfGeminiCursorCLI
Claude CodeCodexWindsurfGeminiCursorCLIVS Code
Source repo————
Safety notes✓Some skills guide Azure or Foundry provisioning, hosted agents, model deployments, toolboxes, workflows, storage, messaging, identity, Key Vault, monitoring, and governance tasks that can affect live cloud resources. Validate resource names, subscriptions, tenants, regions, RBAC, quotas, and costs before approving generated commands or infrastructure changes. The upstream README describes the repository as work in progress with active skill updates and expanding tests. Use skills selectively; loading all skills can dilute context and produce mixed SDK patterns. Skills and MCP configs may invoke external documentation, GitHub, browser automation, Azure CLI, azd, or SDK commands depending on the selected workflow.✓Fabric authoring skills can guide creation or modification of Fabric items, notebooks, schemas, ingestion pipelines, semantic models, reports, Eventstreams, Dataflows Gen2, and deployment automation. Consumption skills may query live Warehouses, Lakehouses, Power BI semantic models, Eventhouse/KQL databases, and catalog metadata; use read-only roles for exploration. Operations skills can investigate performance, health, and slow-query behavior using workspace and workload telemetry; validate before applying generated tuning or remediation steps. Migration and medallion architecture skills can propose broad data-platform changes; review storage paths, costs, retention, governance, and downstream BI impact before execution. The included MCP setup scripts register external Fabric MCP servers only; they do not create, host, or secure a Fabric MCP server for you.✓Wrangler deployment and rollback commands mutate live Worker traffic; verify account, Worker name, route, and environment before running them. The archive URL points to Cloudflare's external Workers SDK tag archive; pin versions and review release provenance before using it in CI. Use least-privilege API tokens and secret stores; never commit `.dev.vars`, `.env`, tokens, account credentials, or CI secret values. Version rollback can restore Worker code/config traffic but does not restore bound KV, R2, D1, Durable Object, queue, or downstream service state. Trigger and route updates can change public reachability; treat them as separate release steps with explicit verification.✓.NET build, test, upgrade, package, template, publish, and migration tasks can modify project files, lock files, generated code, packages, app settings, and deployment artifacts. Diagnostics skills may suggest collecting traces, dumps, counters, crash data, MSBuild binlogs, or performance profiles; collect those artifacts only with explicit approval and storage controls. MCP server skills can expose local code, files, APIs, credentials, or production services as callable tools; review tool descriptions, parameter validation, authorization, and transport choice before connecting clients. NuGet and publish workflows can push packages or artifacts to public or private feeds; verify package IDs, versions, API keys, feed targets, and release policy before publishing. Upgrade and modernization guidance should be verified against each application's framework support window, deployment target, package compatibility, and rollback plan.
Privacy notes✓Azure subscription IDs, tenant IDs, resource names, logs, traces, SDK error output, repository code, architecture notes, and Foundry project details can enter prompts or tool output. Do not paste service principals, access tokens, connection strings, Key Vault secrets, managed identity credentials, or customer data into public issues, screenshots, or committed configs. Microsoft Docs MCP, Context7, GitHub, browser, or other MCP integrations may send query text and project context to external services depending on local configuration.✓Fabric workspace names, tenant IDs, subscription IDs, item IDs, schemas, table names, query text, logs, semantic model metadata, report definitions, and sample data may enter prompts or tool outputs. Do not paste Azure access tokens, Fabric API tokens, connection strings, service principals, workspace secrets, customer data, or regulated datasets into prompts, public issues, screenshots, or committed configs. If you register an external Fabric MCP server, queries and metadata are sent to that server according to its own auth, logging, and retention behavior. Use least-privilege Fabric roles, development workspaces, sample data, or obfuscated datasets when testing assistant-generated Fabric workflows.✓Wrangler deploy uploads bundled Worker source, static assets, and configuration metadata to the target account. Release logs can include request URLs, headers, exception details, route names, account IDs, and zone IDs; redact before sharing publicly. CI deploys rely on API tokens and account IDs; avoid printing environment dumps, secret values, or full request payloads in logs. Worker routes, custom domains, and account metadata can reveal infrastructure topology; keep public notes minimal and intentional.✓.NET repositories may contain connection strings, appsettings secrets, user secrets, certificates, environment variables, telemetry keys, logs, traces, dumps, package credentials, and production data. MSBuild binlogs, crash dumps, profiler output, and test artifacts can contain source paths, dependency graphs, request data, exception payloads, configuration values, and environment details. MCP servers created with these skills may forward prompts and tool inputs to local processes, HTTP services, databases, cloud APIs, or third-party model providers depending on the implementation. Keep private NuGet credentials, signing keys, deployment secrets, customer data, dumps, and proprietary source out of public prompts, issues, pull requests, and shared artifacts.
Prerequisites
  • Node.js and npx for the skills installer.
  • A coding agent or editor setup that can load Agent Skills from `.github/skills`, `.claude/skills`, `.opencode/skills`, or a supported skills directory.
  • Relevant Azure, Microsoft Foundry, Microsoft 365, GitHub, or SDK credentials for the specific skill being used.
  • Current Microsoft Learn, SDK, and package-version checks before implementing against Azure or Foundry APIs.
  • GitHub Copilot CLI installed and authenticated for the recommended plugin marketplace flow.
  • Microsoft Fabric workspace access and the relevant item permissions for Warehouses, Lakehouses, Eventhouses, semantic models, Eventstreams, Dataflows, or reports.
  • Azure CLI and Azure authentication for operations that call Fabric REST APIs or workload-specific endpoints.
  • Workload-specific tokens, endpoints, SQL connection details, KQL endpoints, Power BI permissions, or Spark/Lakehouse context as required by the selected skill.
  • Wrangler-managed Worker project or release plan
  • Project-local Node.js toolchain that satisfies the pinned Wrangler package engines
  • Access to account ID, Worker name, environment name, route, and binding inventory
  • CI or local authentication plan for non-interactive deploys
  • An AI coding assistant or skill host that supports Agent Skills, plugin marketplaces, or direct skill installation.
  • .NET SDK and project-local build/test tooling appropriate for the repository being edited.
  • For MCP server work, the official C# MCP SDK, MCP project templates, and a target transport choice such as stdio or HTTP.
  • For diagnostics, permission to collect traces, dumps, logs, counters, binlogs, or test output from the target environment.
Install
npx skills add microsoft/skills
/plugin install fabric-skills@fabric-collection
npm install --save-dev wrangler@4.97.0 && npx wrangler --version
codex plugin marketplace add dotnet/skills
Config————
Citations
ClaimUnclaimedUnclaimedUnclaimedUnclaimed

Signals

Loading live community signals…

More like this, weekly

A short, calm digest of reviewed Claude resources. Unsubscribe any time.