The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use bash to modify 'config/mcp.json'. This configuration file defines the commands and arguments for MCP servers that the agent subsequently executes as subprocesses, effectively allowing the agent to manage its own executable environment.\n- [COMMAND_EXECUTION]: The 'scripts/evaluation.py' script allows for the execution of arbitrary commands provided via command-line arguments (-c and -a) to test and evaluate MCP server implementations.\n- [PROMPT_INJECTION]: The evaluation harness in 'scripts/evaluation.py' processes untrusted XML data from test files. The content of the 'question' tags is directly interpolated into a system prompt for an LLM without sanitization or boundary markers, creating a surface for indirect prompt injection attacks.\n- [EXTERNAL_DOWNLOADS]: The guide recommends fetching documentation and SDK information from external sources including 'modelcontextprotocol.io' and 'github.com/modelcontextprotocol/*'. These are used to inform the agent's development process and are treated as informative references.

mcp-builder