Loading ToolBench
An MCP server that enables Claude to ask users for input via desktop popups and a web interface. Provides task management with pending/history views, session support, and real-time updates.
Single tool 'AskUser' has a complete input schema with two typed parameters (abstract, content), both with descriptions. However, the tool description is in Chinese and lacks clarity on when/why to use it, error handling, or what the return value represents. No output schema is documented. The tool name lacks a clear action verb (should be 'ask_user' or 'prompt_user'). Parameter descriptions are present but minimal. No error guidance, no idempotency guarantees, no mention of side effects or prerequisites.
调用此工具可以向用户桌面发起一个弹窗，并等待用户回复。
Tool name 'AskUser' violates verb_noun convention. Should be 'ask_user' (lowercase, snake_case) to match LLM parsing expectations and MCP naming standards.
Tool description is in Chinese ('调用此工具可以向用户桌面发起一个弹窗，并等待用户回复。') and lacks English context. Does not explain WHEN to use this tool, what happens on timeout, or what the return value contains. Descriptions must be in English and answer: What? When? Why? What does it return?
No output schema documented. LLMs cannot plan downstream actions without knowing the structure of the reply. Must document that the tool returns a text response and what that response represents.
No error handling guidance. What happens if the user never replies? Timeout? How long? What error is returned? LLMs need explicit recovery paths (e.g., 'If timeout occurs, try again with a shorter content or simpler question').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
Parameter 'content' description is in Chinese and uses vague language ('应优先使用 Markdown 语法'). Should specify: max length, required format, character restrictions, and whether HTML/plain text are acceptable.