Model Context Protocol server for Submail email management system, providing tools to search, read, send, and manage emails with AI-powered features like summarization, reply drafting, and translation
Server has 8 tools with complete input schemas and descriptions in Chinese. Naming follows verb_noun pattern (list_, search_, read_, send_, summarize_, draft_, compose_, translate_). Schemas are well-structured with type constraints, min/max bounds, and format declarations. However, descriptions are in Chinese (non-English), limiting LLM comprehension in English-first contexts. Parameter descriptions are present but terse. Output schemas are not documented. Error handling and recovery guidance are absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Missing idempotency support on write operations.
使用 AI 根据提示词生成邮件内容
使用 AI 草拟邮件回复
列出 Submail 已添加邮箱账号
读取一封邮件正文
按关键词、账号、文件夹、发件人和日期搜索邮件
通过指定账号发送文本或 HTML 邮件,支持 CC、BCC、线程头、幂等键和附件
使用 AI 总结邮件内容
Descriptions are in Chinese; LLMs trained primarily on English will struggle to understand tool purpose and when to invoke them. Descriptions like '列出 Submail 已添加邮箱账号' are not actionable for English-speaking agents.
Output schemas are not documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., what does search_mail return? Does it include message IDs for read_mail?). This forces agents to guess or fail.
No tool annotations (readOnlyHint/destructiveHint/idempotentHint) despite clear risk levels. send_mail is marked WRITE but lacks destructiveHint annotation. Agents cannot distinguish safe retry operations from irreversible ones.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
使用翻译服务翻译邮件或文本
send_mail accepts idempotencyKey parameter but no documentation explains how to use it or what happens on retry. Agents cannot reliably implement idempotent retry logic.
No error handling or recovery guidance. If search_mail returns no results, or send_mail fails due to invalid recipient, agents have no guidance on what to do next (retry? ask user? try alternative?). Error responses must include actionable recovery steps.