Loading ToolBench
智囊团MCP Server - 20位顶级专家的方法论思维。一个MCP服务器，提供来自产品、技术、设计、管理等领域的顶级专家的咨询和圆桌讨论功能。
All 6 tools have descriptions and input schemas, but quality is severely limited. Descriptions are lengthy (200-400+ chars) and written in Chinese narrative style rather than LLM-optimized English. Schemas present but lack proper type definitions for many parameters (e.g., 'advisors' array items have no type constraint). Parameter descriptions are minimal or absent. No output schemas documented. No error handling guidance. Tool names lack action verbs (roundtable, consult, review, brainstorm, critique, suggest are nouns/gerunds, not verb_noun patterns). No tool annotations (readOnlyHint, destructiveHint). Advisor enum values are extensive (40+ IDs) but undocumented in parameter descriptions.
头脑风暴：多位专家从不同角度发散思考，产生创新想法。
向智囊团中的单个专家咨询问题。选择一位专家，获取其专业视角的建议。
设计评审：从多个维度批判性地审视方案，找出问题和改进点。
多角色会诊：邀请多位专家从不同角度评审方案，提供综合建议。
召开圆桌会议：司仪自动分诊（白天桌=事业/技术决策，夜晚桌=人生探讨），点名 3-5 位相关顾问按发言契约轮流发言，最后输出统一格式的会议纪要。用户无需手动挑选专家。
场景建议：根据具体场景和约束条件，给出可执行的方案建议。
Tool names are nouns/gerunds, not action verbs. LLMs infer intent from verb_noun patterns (get_user, create_ticket). Names like 'roundtable', 'consult', 'brainstorm' lack clarity on what action occurs.
Descriptions are lengthy (200-400+ chars) and written in Chinese narrative style. LLM-optimized descriptions should be 50-200 chars in English, stating WHAT, WHEN, and WHY. Current descriptions bury key intent in prose.
Parameter 'advisors' (array type) lacks item type constraint. Schema shows array but no 'items' type definition. Enum values (40+ advisor IDs) are not documented in parameter descriptions, forcing LLMs to guess valid values.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
No output schemas documented. Tools return unstructured responses (meeting notes, advice, suggestions). LLMs cannot plan downstream calls or extract structured data without knowing response field types.
No error handling guidance. Tools marked READ_ONLY but no error recovery patterns documented. No guidance on what happens if advisor lookup fails, context exceeds limits, or LLM generation fails.