The Agent Skills Directory

[SAFE]: The skill provides instructional content for official Microsoft .NET AI libraries, including Microsoft.Extensions.AI and Semantic Kernel. All referenced code and packages are standard for the ecosystem.
[EXTERNAL_DOWNLOADS]: The skill references NuGet packages from Microsoft (e.g., Microsoft.Extensions.AI, Microsoft.SemanticKernel) and links to official Microsoft documentation. These are well-known, trusted sources.
[COMMAND_EXECUTION]: Contains standard .NET CLI commands (dotnet add, dotnet new) for package management and project scaffolding. These commands are typical for development environments.
[CREDENTIALS_SAFE]: Code samples demonstrate best practices for authentication, such as using DefaultAzureCredential from the Azure.Identity package to avoid hardcoded secrets in production code.
[INDIRECT_PROMPT_INJECTION]: The skill provides templates for processing user input through Large Language Models (LLMs). While these implementations are standard for the field, developers using these templates should follow downstream security practices for data sanitization.
Ingestion points: User-provided strings in ChatService.AskAsync and Blazor's SendMessage function.
Boundary markers: None explicitly defined in basic examples.
Capability inventory: LLM chat completion and local tool/function calling.
Sanitization: Not present in the basic documentation snippets; developers are expected to implement this at the application layer.

dotnet-ai