Microsoft announced on 4 November 2026 that it has added installable Microsoft Agent Framework builder skills for .NET and Python. The skills give compatible coding agents guidance for building and modifying Agent Framework applications, including pattern selection, tools, sessions, workflows and hosting.
The distinction for developers is where the skills run. These are development-time skills for a coding assistant. They are not runtime skills loaded by the application that a team ships.
Installation and prerequisites
The skills require a coding agent that supports the Agent Skills format. With GitHub CLI, Microsoft lists gh skill install microsoft/agent-framework agent-framework-builder-dotnet@main for .NET and gh skill install microsoft/agent-framework agent-framework-builder-python@main for Python.
Both commands use main, which gives the coding agent the current repository guidance. Teams that need a reproducible snapshot can pin a tag or commit instead. Microsoft also lists installation through the cross-agent CLI, using npx skills add microsoft/agent-framework --skill agent-framework-builder-dotnet or the equivalent Python skill name.
The installer choice depends on the client and scope supported by GitHub CLI or the skills CLI. The application itself still needs its Agent Framework software development kit packages and model credentials. Installing the builder skill does not remove those requirements.
What the skills tell the coding agent
Each builder skill has a SKILL.md entry point and focused references. The coding agent first inspects the project and its dependencies, then loads the guidance relevant to the requested change rather than treating the whole framework as context for every task.
The guidance is meant to steer implementation choices. It favours the smallest suitable pattern: one agent when tools and sessions are enough, a harness when an operating loop is needed, and a workflow when explicit stages, routing and checkpointing are part of the requirement. Microsoft gives the example that a support assistant does not need a multi-agent system only because it has several tools.
The .NET skill covers AIAgent, Microsoft.Extensions.AI, NuGet packages, dependency injection and ASP.NET Core hosting. The Python skill covers Agent, provider packages, middleware, context providers and Python workflow patterns. Microsoft says the skills also direct the coding agent to clarify consequential unknowns, preserve existing behaviour and validate the change.
Documentation and adjacent Azure work
Microsoft positions the Microsoft Learn Model Context Protocol Server as the companion source for broader documentation. The builder skill supplies focused implementation guidance, while the Learn MCP Server lets the coding agent search Agent Framework documentation and fetch full articles and code examples. Its Streamable HTTP endpoint is https://learn.microsoft.com/api/mcp.
For applications that also use Azure or Foundry services, Microsoft points developers to the Azure Skills Plugin. Its microsoft-foundry skill covers Foundry model and agent workflows, while azure-prepare, azure-deploy and azure-diagnostics address preparation, deployment and troubleshooting for Azure applications.
The plugin includes the Azure MCP Server for tasks such as listing resources and querying logs. It also includes Foundry MCP for model and agent operations. Azure operations require the relevant credentials and permissions.
Prompting and project context
Microsoft's examples show prompts that include the model provider, hosting target, persistence requirements and approval boundaries. A .NET example asks for an Agent Framework support assistant in an ASP.NET Core app, with conversations isolated per user and approval before changing tickets. A Python example asks to turn a research agent into a research-and-review workflow, preserving the response format and adding checkpointing.
That context matters because the builder skill is not a replacement for architecture decisions. It gives the coding agent a framework-specific starting point, but the developer still needs to state provider choices, hosting constraints and approval rules when they are known.
Choosing between a single agent, a harness and a workflow is the kind of agentic design decision ECS's Core AI topic examines, alongside Azure AI Foundry, the EU AI Act and responsible AI. Sessions and workshops give attendees room to compare implementation patterns and discuss where development-time guidance ends and runtime controls begin. Bring concrete project constraints if those decisions are easier to test in conversation than in a coding prompt.
Photo: Lbronn, CC BY-SA 4.0, via Wikimedia Commons, cropped
This article was written by the ECS editorial team with the assistance of Amal Agent, run.events' AI assistant, which supported the research, drafting and proofreading.