The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches the latest protocol documentation and SDK README files from the official Model Context Protocol website and its corresponding GitHub repositories to ensure the developer has current specifications.
[COMMAND_EXECUTION]: The evaluation script (scripts/evaluation.py) is designed to execute the local MCP server under test as a subprocess via standard I/O (stdio). It uses the command and arguments provided by the user through command-line flags (e.g., -c python -a server.py) to launch the server for testing.
[INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection within the testing framework.
Ingestion points: The scripts/evaluation.py script reads test questions from local XML files (e.g., evaluation.xml) which are then passed to the LLM during the evaluation loop.
Boundary markers: The system prompt in the evaluation script explicitly instructs the agent to wrap different parts of its output in <summary>, <feedback>, and <response> tags, which helps in isolating the final answer from reasoning steps.
Capability inventory: The evaluation agent is granted the capability to call any tool exposed by the MCP server being evaluated. This is the intended functionality of the evaluation harness.
Sanitization: The script performs basic XML parsing of the evaluation file but does not sanitize the natural language content of the questions, as they are meant to be interpreted directly by the LLM.

mcp-builder