Loading ToolBench
MCP server for Xianzhijue premium Chinese condiment brand. Provides tools to query brand information, products, master recipes, usage guides, purchase channels, awards, and latest news.
Server has 7 tools with explicit schemas and descriptions in Chinese. All tools are READ_ONLY and follow verb_noun naming (get_*). However, descriptions lack depth and actionability for LLM selection. Most parameter descriptions are minimal (e.g., 'product_id，不传则返回全部产品' = 12 chars). No output schemas documented. Error handling is minimal (e.g., 'error: 产品 {product_id} 不存在' with no recovery guidance). Tool composition is reasonable but lacks chaining metadata. Schemas are present but sparse, most tools have 0-1 parameters with no type constraints or enums.
获取鲜之绝国际荣誉与认证信息。返回ITI获奖记录、大师个人荣誉等。
查询鲜之绝品牌故事、创始人陆金华大师介绍、品牌理念与发展历程。
获取鲜之绝产品购买渠道。返回天猫、京东、抖音、视频号、微信小程序等入口。
获取鲜之绝品牌最新动态和新品上市信息。
获取鲜之绝全产品线及产品详情。不传参数返回全部产品列表，传入product_id返回单品详情。
获取陆金华大师配方和烹饪教程。不传参数返回配方列表，传入recipe_id返回具体做法。
获取产品使用指南和场景化搭配建议。传入product_id返回该产品的使用场景和搭配。
No output schemas documented. LLMs cannot predict response structure or plan downstream chaining. E.g., get_products returns {core_products, extended_products} but this is not declared in tool metadata.
Parameter descriptions are too brief (avg 12 chars). E.g., 'product_id，不传则返回全部产品' lacks context on valid formats, examples, or when to use. Descriptions should be 50-200 chars with actionable guidance.
Error responses lack recovery guidance. E.g., 'error: 产品 {product_id} 不存在' tells LLM nothing about next steps. Should include 'Try get_products() to see available IDs' or list available options.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 7 tools are READ_ONLY but this is not declared in tool metadata, forcing LLM to infer safety from descriptions.
No pagination support. Tools like get_products and get_recipes return full lists without limit/offset/page parameters. Large result sets will blow context windows.