The Agent Skills Directory

[SAFE]: The skill acts as a standard development resource for the Model Context Protocol. The scripts and documentation provided are consistent with legitimate software development and testing practices.
[COMMAND_EXECUTION]: The scripts/connections.py and scripts/evaluation.py files provide functionality to execute local commands using the MCP stdio transport. This allows the harness to launch and interact with a local server process, which is a functional requirement of the protocol and is controlled by the user via command-line arguments.
[INDIRECT_PROMPT_INJECTION]: The testing harness in scripts/evaluation.py possesses an attack surface for indirect prompt injection as it processes external questions to test the MCP server.
Ingestion points: The parse_evaluation_file function in scripts/evaluation.py reads user-supplied XML files containing evaluation questions.
Boundary markers: The EVALUATION_PROMPT uses XML-style tags (<summary>, <feedback>, <response>) to structure and constrain the agent's output.
Capability inventory: The harness can invoke any tool exposed by the MCP server being tested via the call_tool method in scripts/connections.py.
Sanitization: The script interpolates input questions directly into the prompt without specific escaping or sanitization logic.
[EXTERNAL_DOWNLOADS]: The skill's documentation guides the user to fetch official SDK information and protocol specifications from trusted sources such as modelcontextprotocol.io and the modelcontextprotocol GitHub organization.

mcp-builder