MCP server for extracting UI component code from uiverse.io in multiple frameworks (HTML, React, Vue, Svelte, Lit)
This server has severe definition quality issues. Both tools have descriptions in Chinese with low information density and minimal parameter documentation. The parse_and_extract tool has a convoluted description that reads as internal implementation notes rather than a clear agent-facing specification. Input schemas exist but lack proper type annotations and descriptions for parameters. Output schemas are completely undocumented, callers cannot anticipate return structure. No error handling guidance is provided. The list_supported_frameworks tool has a minimal description and no parameters. Neither tool follows the naming or documentation patterns required for production-grade agent tooling.
列出当前支持的框架名称列表(Markdown)。
规则:输入格式为 "框架+空格+链接",例如: HTML https://uiverse.io/Na3ar-17/evil-dragon-24 AI 调用信息: 如果用户输入的链接的开头是 "https://uiverse.io/",那么先识别该前缀后面是否有内容: - 没有内容:不使用 MCP(此处返回说明) - 有内容:根据'框架'选择对应的 MCP 实现提取代码
Input parameter 'query' has no type annotation in schema and the description is cryptic implementation logic ('规则:输入格式为...'), not an LLM-optimized parameter guide. LLMs cannot reliably infer that this tool expects 'FRAMEWORK URL' format or what frameworks are valid.
No output schemas documented for either tool. Callers cannot anticipate structure of parse_and_extract (does it return markdown, raw code, JSON?) or list_supported_frameworks (is it a markdown string or JSON array?). This forces agents to guess and parse fragile string outputs.
No error handling guidance. If parse_and_extract receives an invalid URL or unsupported framework, the code raises ValueError, but there is no guidance in the description about what to do next, what constitutes valid input, or how to self-correct.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 34 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 25 | - | v1 |
Tool descriptions are in Chinese and lack clarity about WHEN to use each tool. The parse_and_extract description reads as pseudo-code implementation notes ('如果用户输入的链接的开头是...') rather than a natural-language guide for LLM decision-making.
Parameter 'query' in parse_and_extract has no enum or pattern constraint. Supported frameworks (HTML, React, Vue, Svelte, Lit) are hardcoded in the implementation but not exposed as schema constraints or parameter descriptions. LLMs will hallucinate invalid framework names.
Naming issue: 'parse_and_extract' violates single-responsibility principle, it both parses input format AND extracts code. This should be split into separate tools or renamed to something like 'extract_uiverse_component' that makes the actual action clear.
The parse_and_extract description embeds example values ('HTML https://uiverse.io/Na3ar-17/evil-dragon-24') which LLMs tend to reuse literally. Should use schema constraints (enum for framework, pattern for URL) instead.