The Agent Skills Directory

[SAFE]: The skill provides technical documentation and reference implementations for the official .NET MCP SDK. No malicious patterns or security risks were identified.\n- [PROMPT_INJECTION]: The skill documents how to handle external data through the MCP protocol, which introduces an attack surface for indirect prompt injection.\n
Ingestion points: Technical handlers for tools, resources, and prompts defined in references/patterns.md receive parameters from external MCP participants.\n
Boundary markers: The documentation in references/security.md recommends specific error handling using McpException to avoid leaking internal system details.\n
Capability inventory: Patterns in references/patterns.md include file system access (File.ReadAllText) and child process execution (StdioClientTransport).\n
Sanitization: The skill demonstrates robust sanitization techniques in references/security.md, including path normalization and root-path verification to prevent traversal attacks.\n- [COMMAND_EXECUTION]: Examples in SKILL.md and references/patterns.md illustrate the use of StdioClientTransport to execute local commands (e.g., npx). These are documented as standard protocol behavior for local tool integration and are accompanied by security advice on process privileges.\n- [DATA_EXFILTRATION]: Code snippets in references/patterns.md demonstrate file reading and network requests. These operations are presented within a security-conscious framework that mandates authorization checks and path validation.

dotnet-mcp