semantic-kernel

The semantic-kernel skill provides guidance for creating, updating, refactoring, explaining, or reviewing solutions built on Microsoft's Semantic Kernel framework, with shared cross-language principles plus dedicated references for .NET and Python. It addresses the problem that Semantic Kernel's API surface evolves quickly, so implementation advice grounded only in model memory becomes stale — the skill repeatedly directs the agent to consult live official documentation and current samples before making recommendations or code changes.

The workflow starts by determining the target language: the .NET path applies when the repository contains C#/.NET project files or the user asks for C#, and the Python path applies for `.py`, `pyproject.toml`, or `requirements.txt` projects, with tie-breaking rules for mixed or ambiguous repositories. Shared guidance covers using async patterns for kernel operations, following official plugin and function-calling patterns, implementing explicit error handling and logging, preferring strong typing and maintainable composition, using built-in connectors for Azure AI Foundry, Azure OpenAI, and OpenAI (favoring Azure AI Foundry for new projects), leveraging the kernel's memory and context management, and using DefaultAzureCredential where Azure authentication is appropriate. The language reference files point to the authoritative upstream repository and samples for each ecosystem and give language-specific practices. Completion criteria ensure recommendations, package names, and sample paths match the selected ecosystem and current documentation.

The target users are developers building Semantic Kernel applications, plugins, function-calling flows, or AI integrations in either .NET or Python. It is a documentation-and-conventions reference that promotes secure defaults such as managed-identity authentication via DefaultAzureCredential, and contains no executable or risky behavior of its own.