The Agent Skills Directory

[SAFE]: The skill provides legitimate developer documentation and utility scripts for the Model Context Protocol. No malicious patterns, such as unauthorized data exfiltration, credential theft, or persistence mechanisms, were detected.
[COMMAND_EXECUTION]: The evaluation harness (scripts/evaluation.py) is designed to launch MCP servers via the stdio transport. This involves executing local commands and arguments provided by the developer through the command-line interface. This is a core requirement for testing local MCP integrations.
[INDIRECT_PROMPT_INJECTION]: The evaluation.py script ingests test questions from user-provided XML files and interpolates them into a system prompt for an LLM. While this is a common surface for indirect prompt injection, it is mitigated by the tool's intended use as a local development utility where the user controls the input XML file.
[EXTERNAL_DOWNLOADS]: The documentation encourages fetching official MCP SDK documentation and specifications from trusted sources like GitHub (modelcontextprotocol organization) and the official protocol website. These are standard resources for developers.

mcp-builder