Microsoft Agent Skills: Context-Driven Development for Azure AI Agents

Forum topic · ✨步子哥 · 2026-01-27

Summary

Microsoft's open-source agent-skills repository provides modular, context-driven skill packs that transform generic AI coding assistants into Azure and Microsoft Foundry domain experts. The post introduces the core problem of "context rot"—when overloaded context windows dilute model attention and degrade output quality—and explains the repo's rule of loading only 1-3 relevant skills per project. It walks through the repository structure (.github/skills/, prompts/, agents/, workflows/, MCP configuration) and catalogs available skills including Azure-hosted agent frameworks, Azure AI Agents SDK, AZD deployment with Bicep, Azure AI Search, Voice Live, Cosmos DB, FastAPI, Foundry IQ, NextGen frontend, GitHub issue automation, MCP server building, podcast generation, Pydantic modeling, React Flow, and Zustand. Advanced strategies for large codebases (Graph RAG for code, AST-based memory, context graphs) and a four-step quickstart are also covered. Everything is MIT licensed.

This post introduces Microsoft's open-source agent-skills repository (https://github.com/microsoft/agent-skills), framed as a "context-driven development" approach: AI agents produce quality code only when fed precise, relevant context.

Context-driven development architecture

The Core Warning: Context Rot

The author highlights a critical pitfall called context rot: when an AI's context window is filled with irrelevant or outdated information, model attention is diluted, tokens are wasted, and output quality drops sharply—causing hallucinations or confusion between frameworks.

> Never load all skills at once. Select only the 1–3 skills most relevant to the current project.

Repository Layout

.github/skills/ — the skill library; each subfolder is an independent knowledge pack (SKILL.md)
prompts/ — reusable prompt templates (code review, new components, API endpoints)
agents/ — AI persona definitions (.agent.md: backend expert, frontend designer, planner)
.vscode/mcp.json — Model Context Protocol server configuration
workflows/ — automation, e.g., weekly Foundry docs sync

Available Skills

agent-framework-azure-hosted-agents — persistent, hosted agents on Azure via the Microsoft Agent Framework SDK
azure-ai-agents-python — low-level SDK: agent CRUD, thread management, streaming, tool registration
azd-deployment — one-click deployment to Container Apps with Azure Developer CLI + Bicep
azure-ai-search-python — vector, hybrid search, and agentic retrieval
azure-ai-voicelive-skill — real-time voice interaction for agents
cosmos-db-python-skill — Cosmos DB NoSQL CRUD patterns with FastAPI and partition strategies
fastapi-router — routing, auth, and Pydantic response model best practices
foundry-iq-python — knowledge-base-driven agentic retrieval
foundry-nextgen-frontend — Vite + React UI patterns per Microsoft's latest design language
foundry-sdk-python — versioned agents, evaluation metrics, production features
issue-creator — standardized GitHub issue creation, labeling, PR linking
mcp-builder — build MCP servers in Python or Node (Playwright, Terraform, etc.)
podcast-generation — full script-to-audio multi-agent workflow
pydantic-models — Base/Create/Update/Response layered model design
react-flow-node — draggable interactive flow nodes with TypeScript and Zustand
zustand-store — lightweight global state management

Sharing Across Projects

Reuse a skill in multiple projects via symlinks (examples provided for macOS, Linux, and Windows).

Strategies for Large Codebases

1. Graph RAG for Code — semantic search over functions, classes, and module relationships 2. AST-based Memory Systems — hierarchical abstract syntax tree maps 3. Context Graph — knowledge graphs of component dependencies, call relationships, and data flow

Always-Fresh Foundry Docs

The repo is indexed by Context7 (https://context7.com/microsoft/agent-skills) with a weekly auto-sync workflow, so MCP-connected agents can query up-to-date official documentation.

Quickstart (4 Steps)

1. Copy only the skills you need into .github/skills/ 2. Configure your agent to load them (VS Code Copilot auto-discovers; Claude requires manual reference) 3. Copy .vscode/mcp.json to enable tool servers (GitHub, Playwright, Terraform, Microsoft Docs, Sequential Thinking, Memory) 4. Optionally add Context7 for Foundry docs

Licensing & Contribution

The project is MIT licensed and welcomes community contributions of new skills, prompt improvements, and MCP configurations.

References

1. https://github.com/microsoft/agent-skills 2. https://devblogs.microsoft.com/all-things-azure/context-driven-development-agent-skills-for-microsoft-foundry-and-azure/ 3. https://context7.com/microsoft/agent-skills 4. Azure AI Foundry Official Documentation (via Context7 sync) 5. Model Context Protocol (MCP) Specification and Tool Integrations