The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Python evaluation script (scripts/evaluation.py) and a connection utility (scripts/connections.py) that facilitate the execution of local MCP servers via the stdio transport. This functionality allows the user to launch their own server processes for testing, which is the standard operational mode for the Model Context Protocol.
[EXTERNAL_DOWNLOADS]: The instructions and reference guides utilize WebFetch to retrieve official MCP documentation and SDK README files from modelcontextprotocol.io and github.com/modelcontextprotocol. These sources are the authoritative references for the protocol and its development libraries.
[PROMPT_INJECTION]: The evaluation script processes external data from XML files and tool outputs from the MCP server being tested. It employs structured output tags (<summary>, <feedback>, <response>) to maintain agent behavior during the evaluation loop. This is a common pattern for automated testing of LLM capabilities and does not show malicious intent.
[CREDENTIALS_UNSAFE]: The skill correctly implements security best practices for credential management. The scripts and instructions consistently recommend using environment variables (e.g., ANTHROPIC_API_KEY, GITHUB_TOKEN) rather than hardcoding secrets, preventing accidental exposure.

mcp-builder