The Agent Skills Directory

[COMMAND_EXECUTION]: The evaluation utility (scripts/evaluation.py) launches local MCP server processes using the standard mcp library's stdio_client to perform automated testing.
[EXTERNAL_DOWNLOADS]: The developer guide (SKILL.md) recommends fetching protocol specifications and SDK documentation from official Model Context Protocol domains and GitHub repositories.
[DATA_EXFILTRATION]: The testing script (scripts/evaluation.py) communicates with the Anthropic API to process questions and transmit tool invocation data during evaluations.
[PROMPT_INJECTION]: The evaluation harness (scripts/evaluation.py) reads data from a user-supplied XML file to generate prompts for an LLM.
Ingestion points: Questions are ingested from an XML file provided by the user via the eval_file argument.
Boundary markers: The system prompt in scripts/evaluation.py employs XML tags (<summary>, <feedback>, <response>) to structure the agent's output.
Capability inventory: The script is capable of executing tools exposed by the target MCP server and writing evaluation reports to the local file system.
Sanitization: Content from the evaluation file is processed without explicit sanitization or filtering before being sent to the LLM.

mcp-builder