A modular MCP server system consisting of a proxy server that manages multiple downstream MCP servers (camera capture, image generation, Google search, weather query) and a client that interfaces with them via DashScope LLM.
This server has severe definition quality issues. Tool names lack clarity and verb-based naming conventions. Descriptions are present but often underdeveloped or in Chinese without consistent English translations. Input schemas exist but lack proper type definitions in several cases, and parameter descriptions are minimal. No output schemas are documented. Error handling is present but minimal. The proxy_tool_call tool is particularly problematic, it accepts arbitrary tool names and args, making validation impossible. Overall, the definition quality falls well below production standards.
使用摄像头拍照并分析微表情
使用ComfyUI生成图片。 :param prompt: 正向提示词(描述想要生成的图片内容) :param negative_prompt: 负向提示词(描述不想要的元素,默认:text, watermark) :param width: 图片宽度(默认:512) :param height: 图片高度(默认:512) :return: 生成的图片保存路径或错误信息
使用 Selenium 执行 Google 搜索,返回前 10 个非广告结果的标题、链接和摘要。 Args: query (str): 搜索关键词,例如 'Python tutorial' Returns: str: 前 10 个非广告搜索结果的标题、链接和摘要
代理工具,根据工具名动态调用其他服务端的工具,输入格式为字典:{'tool': 'tool_name', 'args': {...}}
输入高德地图城市代码,返回今日天气查询结果。 :param city_code: 高德地图城市代码(例如北京是 '110000') :return: 格式化后的天气信息
proxy_tool_call lacks verb-based naming and has dangerously generic schema, accepts any tool name and arbitrary object args, preventing validation and making LLM tool selection unreliable
capture_camera_image has empty input schema ({}), making it impossible for an LLM to understand required parameters; description is vague in Chinese without English equivalent
No output schemas documented for any tool, LLMs cannot infer what fields to expect or plan downstream calls; responses appear to be unstructured strings
Tool descriptions contain Chinese text without consistent English translations; inconsistent description quality across tools (some 20 chars, others 200+)
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 34 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 44 | - | v1 |
Parameter descriptions are minimal or missing context, e.g., 'query' in google_search lacks examples of valid query formats; 'city_code' in query_weather assumes knowledge of Amap city codes without explanation
No error handling guidance, tools return plain error strings like '⚠️ 工具调用失败' without actionable recovery steps for LLMs
Tool names lack action verbs, 'proxy_tool_call' and 'capture_camera_image' and 'generate_image' are vague; should be 'call_remote_tool', 'capture_facial_expression' or similar for clarity