creating-glaido-mcp-servers

SUSPICIOUS. The skill’s purpose is coherent, and most behavior fits building local Glaido MCP servers, but it materially increases trust surface by telling the agent to discover, install, and run third-party MCP servers and runtimes from the web. The main risk is supply-chain and credential forwarding into externally sourced local servers, not confirmed malware or hidden exfiltration in this skill itself.

The code is a readable MCP configuration validator. It does not show evidence of malware, data exfiltration, persistence, obfuscation, or credential theft. Its significant security property is that it launches the command specified by the inspected mcp.json, which can execute arbitrary code if the configuration is untrusted. This is an inherent risk of the launch-check feature rather than concealed malicious behavior; use --no-launch when inspecting untrusted projects.