Agent 365 Skills: Bring your agents into Microsoft Agent 365 in minutes

Your agent works beautifully on your laptop. It reasons, calls tools, and nails the demo. Then comes the part nobody enjoys: making it enterprise-ready. Identity. Observability. Governance. Messaging. Secure access to Microsoft 365 data. Until now, that last mile meant reading through Microsoft Learn articles, running CLI commands in exactly the right order, hand-editing manifests, and hoping every prerequisite was in place before the agent reached production.

Organizations are moving fast — from experimenting with AI agents to putting them to work across real business processes. Developers are building with LangChain, OpenAI, Semantic Kernel, Azure AI Foundry, Claude SDK, Google ADK, and more. But great orchestration logic isn’t enough. Agents also need identity, observability, governance, messaging, testing, and secure access to Microsoft 365 data.

Agent 365 Skills collapse that workflow into a guided, natural-language experience inside the coding assistant developers already use. Describe what you need — “add observability,” “wire up WorkIQ Mail and Calendar,” “test this agent locally” — and the skill detects your project, asks only the questions it has to, applies the right changes, and validates the work before the session ends.

The last mile is the hard mile

Before Agent 365 Skills, onboarding an agent often meant a long, fragile chain of manual steps: installing the Agent 365 CLI, validating Azure prerequisites, registering a blueprint, configuring Entra artifacts, wiring OpenTelemetry exporters, packaging manifests, connecting MCP servers, and testing locally. Miss one step, and things break — often late, in production.

From a multi-hour, order-sensitive gauntlet to one guided sentence.

Agent 365 Skills replace that disconnected workflow with a guided lifecycle. Developers stay in Claude Code, GitHub Copilot CLI, or VS Code agent mode and ask for the outcome they want. The skill handles the implementation details — while preserving the technical controls enterprise teams depend on.

What changes for developers:

• Start with one sentence — “add WorkIQ tools” or “integrate observability” — instead of reading many Microsoft Learn pages.
• Auto-instrument in minutes instead of manually wiring OpenTelemetry exporters, token resolvers, and baggage scopes.
• Let the skill apply the required manifest and configuration updates instead of hand-editing them.
• Catch issues before shipping with built-in validators instead of hoping nothing was missed.

Watch it work

Say a developer wants telemetry. They type a single line:

> add observability to this agent

From there, the skill takes over. It detects the framework — say, a .NET Semantic Kernel project — asks only the questions that matter (which authentication mode: OBO, Agentic-User, or S2S), wires OpenTelemetry and the Agent 365 tracing exporter, and confirms the project still builds. No manifest spelunking. No half-configured exporters.

Every skill follows the same five-beat loop: describe, detect, ask, apply, validate.

Other requests work the same way:

• “Add messaging and notifications to this agent.”
• “Add observability to this agent.”
• “Wire up WorkIQ Mail and Calendar.”
• “Test this agent locally.”

Six skills, one lifecycle

Agent 365 Skills include six skills that map to the lifecycle steps developers hit when preparing an agent for Microsoft Agent 365. Run only the ones your agent needs — or use the make skills to execute all the required steps.

Built for the stack you already have

Agent development isn’t one language or one framework, and Agent 365 Skills are designed for that reality. They support three major ecosystems and many of the most popular agent frameworks.

• .NET: Agent Framework, Semantic Kernel, and Azure AI Foundry.
• Node.js: LangChain, OpenAI Agents SDK, Claude SDK, Semantic Kernel, and Google ADK.
• Python: Agent Framework, LangChain, OpenAI, Claude, Semantic Kernel, and Google ADK.

The project stack is auto-detected, so you don’t start by writing configuration files or mapping your project by hand. The skill determines the path and guides the next step.

Enterprise-ready by design

Speed matters — but so does not breaking things. Agent 365 Skills are built to move fast without giving up the safety and validation teams expect.

• Additive and idempotent: Skills don’t delete or restructure existing code, and re-running a skill is safe by design.
• State-aware: A workspace detection cache tracks what’s already wired, so completed steps are skipped, not duplicated.
• Guard-railed: Pre-tool hooks block modifications outside the project, and stop-hook validators verify the build passes before the session ends.

Together, these controls reduce setup risk while giving developers a repeatable path to register agents, add observability, connect Microsoft 365 data, and test locally.

What your agent gains

One onboarding flow connects your agent to identity, observability, data, messaging, and testing.

For customers, Agent 365 Skills shorten the path from prototype to enterprise-ready agent. Developers spend less time stitching setup steps together and more time building agents that solve business problems. IT and security teams get more consistent registration, observability, and governance patterns across the agent estate.

The result is a faster, safer onboarding experience: agents registered with Agent 365, instrumented for visibility in Defender, Purview, and the Microsoft 365 admin center, connected to Microsoft 365 data through Work IQ MCP servers, and tested locally before deployment.

Get started

Install the skills with:

gh skill add microsoft/agent365-skills

• Explore the GitHub repository.
• Read the Microsoft Agent 365 developer docs for deeper developer guidance.