mcp-factory is a manifest-driven MCP server scaffolder and runtime hub, not a production tool provider. The 4 tools present are either demo/fixture tools or test utilities. Tool definitions lack proper schemas, have minimal descriptions (under 50 chars), and parameters are either missing or lack type information. The say_hello tool is explicitly marked as a demo with a disclaimer. The ping, echo, and crash tools are from test fixtures, not production-grade integrations. No parameter type definitions are visible in the source. Descriptions are generic and do not meet the 10 - 1024 character expectation for clarity. Error handling is minimal. This server's purpose is scaffolding and hub management, not exposing user-facing tools, the low score reflects that the tools are illustrative rather than production-quality.

• Expand ping description from 'Ping the server' to a 50 - 150 character explanation: 'Check server health and connectivity. Returns 'pong' if available. Use to verify the hub is responding before other operations.'
• Add explicit error recovery guidance to crash tool: 'Always returns an error (diagnostic tool). If encountered unexpectedly, verify hub logs for system issues or restart the server.'
• Document say_hello's actual behavior separately from its demo-mode disclaimer. Add: 'Greet a user by name. Returns a welcome message. Useful for testing basic tool invocation and verifying agent connectivity.'
• Add parameter type and validation to echo tool: 'message (string, 1 - 10000 chars): The text to echo back. No formatting applied. Useful for testing message routing and content preservation.'
• Provide explicit output schemas for all tools. E.g., ping returns {status: string, timestamp: string}, echo returns {message: string}, crash returns {error: string}.
• If these are truly test fixtures, move them out of the tool roster in production. The demo tools should only appear when MCP_FACTORY_NO_DEMO is not set, and should be documented as such in the server capabilities.