agent-framework
Build AI agents with Microsoft Agent Framework
Create AI agents and multi-agent workflows using Microsoft Agent Framework SDK. This skill scaffolds agent projects with tools, HTTP server support, and VSCode debugging configurations for enterprise AI applications.
Download the skill ZIP
Upload in Claude
Go to Settings → Capabilities → Skills → Upload skill
Toggle on and start using
Test it
Using "agent-framework". Create a new agent called 'WeatherAgent' that can tell the weather
Expected outcome:
The skill generates: requirements.txt with pinned agent-framework versions, main.py with ChatAgent setup and weather tool, .env file template for FOUNDRY_PROJECT_ENDPOINT and FOUNDRY_MODEL_DEPLOYMENT_NAME, .vscode/launch.json and .vscode/tasks.json for debugging
Using "agent-framework". Build a Writer-Reviewer workflow
Expected outcome:
The skill generates: A Writer executor that creates content, A Reviewer executor that provides feedback, WorkflowBuilder configuration with edges connecting the two agents, run_stream event handling for real-time output
Security Audit
SafeThis is an official Microsoft skill for creating AI agents using Microsoft Agent Framework SDK. The static scanner flagged 165 potential issues, but evaluation confirms all are false positives. The flagged backtick patterns are Markdown code fences in documentation files. The skill appropriately uses .env files for storing Azure Foundry credentials (FOUNDRY_PROJECT_ENDPOINT, FOUNDRY_MODEL_DEPLOYMENT_NAME), which is a standard and secure practice for credential management.
Medium Risk Issues (1)
Risk Factors
🔑 Env variables (1)
Quality Score
What You Can Build
Enterprise AI Application Development
Build production AI agents with type safety, checkpointing, and orchestration for enterprise Azure environments.
Multi-Agent Workflow Prototyping
Create and test multi-agent workflows with Writer-Reviewer patterns, loops, and human-in-the-loop interactions.
VSCode Debugging Setup
Configure local debugging with AI Toolkit Agent Inspector for interactive agent testing and troubleshooting.
Try These Prompts
Create a new AI agent using Microsoft Agent Framework that can respond to user queries. Include HTTP server mode support.
Build an agent with custom tools using the function calling pattern. Add at least 2 custom tools for the agent to use.
Create a multi-agent workflow with a Writer agent that generates content and a Reviewer agent that provides feedback. Use the workflow builder pattern.
Build a multi-agent loop workflow where a Teacher agent and Student agent interact. Include turn-based control with a maximum of 5 iterations.
Best Practices
- Always use pinned SDK versions (1.0.0b260107 for agent-framework) to avoid breaking changes from beta releases
- Use virtual environments (.venv) to isolate dependencies and avoid conflicts
- Configure HTTP server mode as the default entry point for better debugging and containerization support
- Pin environment variables in .env file but remind users to update values before running
Avoid
- Using bare 'python' or 'pip' commands without virtual environment activation
- Skipping the server mode when the user needs HTTP endpoint access
- Using legacy AzureAIAgentClient instead of current AzureAIClient
- Not pinning SDK versions leading to breaking changes from beta releases