The Agent Skills Directory

[COMMAND_EXECUTION]: The evaluation harness (scripts/evaluation.py) and connection logic (scripts/connections.py) are designed to execute local commands to launch MCP servers via standard input/output (stdio). This is a core functionality for developers to test their implementations locally.
[EXTERNAL_DOWNLOADS]: The skill references and fetches documentation from official sources, specifically modelcontextprotocol.io and the modelcontextprotocol organization on GitHub. These are well-known resources for the Model Context Protocol.
[PROMPT_INJECTION]: The evaluation harness in scripts/evaluation.py processes XML files containing test questions which are passed directly to an LLM without specific boundary markers. While this is an indirect prompt injection surface, it is consistent with the intended use case of an automated evaluation tool.
Ingestion points: scripts/evaluation.py reads data from the XML evaluation file provided via the command line.
Boundary markers: Absent; the question text is interpolated directly into the user message content.
Capability inventory: The evaluator agent has access to tools provided by the MCP server being tested, which may include file system, network, or subprocess capabilities depending on the server implementation.
Sanitization: Absent; the script only performs basic whitespace stripping on the input text.

mcp-builder