The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches protocol specifications and SDK documentation from official sources, including modelcontextprotocol.io and the modelcontextprotocol GitHub organization.
[INDIRECT_PROMPT_INJECTION]: The evaluation harness processes external XML files containing test cases that are passed to the agent during evaluation.
Ingestion points: Test questions are parsed from user-supplied XML files in scripts/evaluation.py via parse_evaluation_file.
Boundary markers: The script employs a system prompt (EVALUATION_PROMPT) that enforces the use of structured XML tags (<summary>, <feedback>, <response>) to separate analysis from results.
Capability inventory: The script is capable of calling any tool exposed by the MCP server being evaluated via the connection.call_tool method.
Sanitization: The questions are passed to the LLM as parsed from the XML without additional text sanitization or filtering.
[COMMAND_EXECUTION]: The evaluation script facilitates the execution of local server commands (e.g., python server.py) when using the stdio transport mechanism. This is a primary feature designed to allow developers to test their local implementations.

mcp-builder