The Agent Skills Directory

[SAFE]: The skill serves as documentation for the official MCP C# SDK and associated NuGet packages. All links and repository references point to official and reputable sources.
[COMMAND_EXECUTION]: The documentation includes patterns for launching local MCP servers via shell commands (e.g., using npx). These are standard integration examples for the protocol and do not contain malicious payloads.
[REMOTE_CODE_EXECUTION]: Code examples illustrate connecting to remote servers or running packages from the official @modelcontextprotocol registry. These are intended for development workflows and use trusted namespaces.
[CREDENTIALS_SAFE]: Authentication examples use safe placeholders like <token> and advise on implementing security at the transport boundary (ASP.NET Core middleware), adhering to secret management best practices.
[INDIRECT_PROMPT_INJECTION]: While the skill describes an interface for processing external data (MCP tools/resources), it provides mandatory security guidance to mitigate risks:
Ingestion points: Arguments passed to tools and resource URI templates defined in references/patterns.md.
Boundary markers: Encourages explicit capability negotiation and validation of all incoming parameters.
Capability inventory: File system access, network transport (stdio, HTTP), and chat client integration patterns.
Sanitization: Explicitly provides code examples for path normalization and validation (fullPath.StartsWith(root)) to prevent path traversal attacks.

mcp