Loading ToolBench
A Model Context Protocol server that provides access to prompt templates stored in a Notion database. It allows listing available prompts as resources, reading individual prompts, and composing final prompts by combining templates with user input.
Server has 9 tools with basic schemas and descriptions, but significant quality gaps. All tools are READ_ONLY and descriptions are in Chinese (non-standard for LLM optimization). Schemas present but minimal, most parameters lack detailed constraints, ranges, or format guidance. No parameter descriptions beyond bare names. Error handling is absent from visible code. Tool names follow verb_noun pattern (positive), but descriptions are generic and under-optimized for LLM selection. No output schema documentation visible. Composition is reasonable (single responsibility per tool), but parameter design lacks depth expected of production tools.
将用户输入整合到提示词模板中
通过名称获取提示词
获取特定类别的提示词
列出所有可用的提示词类别
列出所有可用的提示词
依次处理指定类别的所有提示词，每个提示词单独处理
将用户输入整合到提示词模板中，并使用当前LLM处理
Descriptions are in Chinese and generic. LLM optimization requires English, 50 - 200 character descriptions that explain WHAT, WHEN, and WHY. Current descriptions like '列出所有可用的提示词' (list all available prompts) lack context for tool selection and are not optimized for English-speaking LLMs.
Parameter descriptions are missing or minimal. 'name' in get_prompt_by_name has description '提示词名称' (prompt name) but lacks format guidance, constraints, or examples. Parameters need 50 - 100 character descriptions explaining expected format, valid values, and dependencies.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | <=2025-11-25 | v2 |
刷新提示词缓存
搜索提示词（名称、描述和内容）
No output schema documentation. Tools return results but no visible schema definition for response structure. LLMs cannot plan downstream calls or extract fields without knowing what fields are returned. E.g., does list_prompts return [id, name, description, category, content]? Unknown.
No error handling guidance visible. Code does not show how tools handle missing prompts, invalid categories, or API failures. Error responses should guide LLM recovery: 'Prompt not found. Try search_prompts() with a partial name.' Current code throws generic errors.
Ambiguous tool responsibilities. 'process_composed_prompt' and 'compose_prompt' are similar, one composes, one composes AND processes. LLMs will conflate them. Clarify: does compose_prompt return a string, or does it invoke an LLM? Naming and descriptions must disambiguate.