The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/connections.py and scripts/evaluation.py scripts facilitate the execution of local processes through the stdio transport. This capability is required for the skill's primary purpose of testing MCP servers in a development environment.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs the agent to retrieve protocol specifications and SDK information from official documentation sites and public GitHub repositories. These references target well-known technology providers and do not involve executable code from untrusted sources.
[INDIRECT_PROMPT_INJECTION]: The evaluation harness in scripts/evaluation.py processes output from external MCP tools and provides it to the LLM. This creates a surface for indirect prompt injection if the tool output contains adversarial content. The skill mitigates this risk by using XML tags to delimit untrusted content and system instructions that explicitly warn the model to treat tool output as data rather than instructions.
[DYNAMIC_EXECUTION]: The evaluation harness dynamically loads and interacts with external tool schemas and server executables at runtime to verify their protocol compliance.

mcp-builder