The Agent Skills Directory

[SAFE]: The skill is purely instructional, providing architectural guidance and best practices for creating MCP servers without including any malicious scripts or hidden commands.
[EXTERNAL_DOWNLOADS]: The skill fetches documentation from the official Model Context Protocol website and its public GitHub repositories (python-sdk and typescript-sdk). These references are used for reading technical specifications and do not involve executing remote code.
[COMMAND_EXECUTION]: Instructions include standard development commands for testing local servers, such as using timeout or python for execution. These are presented as educational examples for manual developer use.
[DATA_EXPOSURE]: The skill explicitly promotes secure design by recommending input validation (Pydantic/Zod), truncation of large responses to protect context windows, and the creation of actionable but safe error messages.

mcp-builder