Loading ToolBench
基于Model Context Protocol(MCP)协议开发的AI辅助工具，旨在帮助开发者快速构建软件原型。
This MCP server has fundamental definition quality issues across all 5 tools. Tool naming lacks consistent action verbs (hello_world, getSpec, init are non-standard). Descriptions vary wildly in quality: some are in Chinese with operational dependencies documented (start, stop), others are generic single-liners (hello_world: 'A simple hello world tool'). Input schemas are present but minimal, most tools have empty schemas {} with no parameter details. No output schemas are documented. Error handling is partially present (webTool.ts shows error returns) but lacks recovery guidance. The server demonstrates basic MCP understanding but falls short of production-grade tool design.
获取原型创建的规范说明文档
A simple hello world tool that returns a greeting message
初始化MCP服务，设置原型根目录和服务端口
该工具将启动web服务并展示模型，启动后会后台运行，所以避免重复调用！ 调用成功后，请在浏览器里输入本地地址和 init 中指定的端口号进行查看。 在调用此工具前，下面的事情需要就位，否则此工具可能无法正常工作： - 遵循 getSpec 工具中的指示并进行了原型创造。 - 已经调用 init 工具进行了初始化。
停止展示原型，并关闭web服务器
Tool naming does not follow verb_noun convention. 'hello_world' and 'getSpec' lack action verbs; 'init' and 'start' are ambiguous about what they initialize/start. Recommended: rename to 'send_greeting', 'get_specification', 'initialize_service', 'launch_web_server'.
Most tools (getSpec, start, stop) have EMPTY input schemas {} with NO parameter documentation. This prevents LLMs from understanding valid parameters.
No output schemas are documented for any tool. LLMs cannot know what fields to expect or plan downstream calls. A tool returning a greeting, specification doc, or web server status needs clear response structure definition.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 35 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 28 | - | v1 |
Descriptions are inconsistent in language (English vs Chinese) and detail level. 'hello_world' has 63 chars (acceptable), but 'getSpec' and 'stop' are sparse. Start/stop tools include operational prerequisites in Chinese that should be in English for international LLM agents.
'init' tool accepts prototypeRoot and port parameters but descriptions lack validation constraints. Is port required? What is the valid range (1-65535)? What is the expected format for prototypeRoot (absolute path, relative, must exist?).
Tools 'start' and 'stop' modify state (launch/terminate web server) but error handling lacks recovery guidance. webTool.ts returns generic error objects; LLMs cannot infer 'Should I retry? Preconditions?'
Tool 'hello_world' requires a 'name' parameter but lacks example constraints. Should LLMs pass long strings? Emoji? Null? Description should specify expected format and length bounds.