Loading ToolBench
MCP 服务器，用于管理桌面图片，提供图片文件统计、查看和压缩功能
Three tools with basic schemas and Chinese descriptions. Tool naming follows verb_noun pattern (count-, list-, compress-) but descriptions are minimal (10-20 chars in English translation). Schemas present for compress-image but empty for the first two tools. No output schema documentation. Error handling exists but lacks recovery guidance. No parameter descriptions for count/list tools. Missing idempotency and composition guidance.
压缩图片
统计桌面上的图片文件数量
获取桌面上的图片文件名称列表
Two tools (count-desktop-images, list-desktop-images) have empty input schemas ({}).
All tool descriptions are under 20 characters (Chinese: '统计桌面上的图片文件数量' ~12 chars, '获取桌面上的图片文件名称列表' ~15 chars, '压缩图片' ~4 chars). Descriptions lack WHAT/WHEN/WHY context needed for LLM tool selection.
No output schema documented for any tool. LLMs cannot predict response structure or plan downstream calls. compress-image returns text with file sizes; count/list return text counts/lists, all undocumented.
compress-image accepts quality parameter (1-100) with default 80, but no guidance on what quality values mean or when to use lower/higher values. Parameter description is minimal.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | 2026-07-28+ | v2 |
Error responses return raw error messages but no recovery guidance. E.g., 'File not found' does not suggest calling list-desktop-images first. Errors lack actionable next steps for the LLM.