The Agent Skills Directory

[COMMAND_EXECUTION]: The evaluation scripts (scripts/evaluation.py and scripts/connections.py) facilitate the execution of arbitrary local commands through the stdio transport. This is a core feature for testing local MCP servers but allows for the execution of any system command provided as an argument.\n- [EXTERNAL_DOWNLOADS]: The documentation references fetching technical specifications and SDK details from modelcontextprotocol.io and the modelcontextprotocol GitHub repositories. These are recognized as well-known sources for the Model Context Protocol.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). The automated evaluation loop processes external questions from XML files and outputs from testing tools. Ingestion points: scripts/evaluation.py reads user-provided XML files and receives tool results from MCP servers. Boundary markers: The EVALUATION_PROMPT uses XML tags to structure the agent's output but lacks markers to isolate ingested content. Capability inventory: scripts/connections.py supports subprocess execution and network requests. Sanitization: No explicit content sanitization or instructions to ignore embedded commands are present.

mcp-builder