This server has 3 tools with minimal quality. Tool names follow verb_noun convention (get_, save_), but descriptions are in Chinese and very brief (12-26 chars in English equivalent), violating the 50-200 char guideline for LLM optimization. Input schemas are present but sparse: get_user_info has empty schema {}, while save_markdown and save_html have partial schemas with only 2 parameters each. No parameter descriptions are visible in the schema (the 'description' field shows only generic Chinese text like '对话内容' = 'conversation content'). No output schemas are documented. Error handling exists but does not guide LLM recovery, exceptions are raised with generic messages like 'HTTP request failed' without actionable next steps. Tool descriptions do not state what data is returned or when to use each tool vs alternatives. The server accepts API credentials via environment variable (good security practice) but does not mention this in tool descriptions, leaving ambiguity about prerequisites.
获取当前用户信息
保存 html 代码到 LimeAI
保存文本内容(markdown格式)到 LimeAI
get_user_info has empty input schema ({}), no schema documentation visible. Per hard scoring rule, schema score must be 0.
All tool descriptions are extremely brief (Chinese originals; English equivalent 12-26 chars). Baseline for A+ tools is 50-200 chars. Descriptions must state WHAT the tool does, WHEN to use it, and what it RETURNS. Current descriptions ('Get current user info', 'Save markdown content to LimeAI') lack return type, prerequisites, and when to call instead of alternatives.
save_markdown and save_html parameter descriptions in schema are minimalist Chinese ('对话内容' = 'conversation content', '文件名' = 'filename'). LLMs cannot infer meaning from generic translations. Descriptions must explain expected format, constraints (e.g. max length), and whether the param is optional and what the default is.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 32 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 36 | - | v1 |
No output schemas documented. LLMs cannot know what fields to expect in responses, breaking tool chaining and downstream planning. Even a brief 'Returns: {status: string, message: string, data: object}' in the description would help.
Error handling does not guide LLM recovery. Errors like 'HTTP request failed: ...' or 'API response error: ...' tell the agent nothing about whether to retry, ask the user, or give up. No error classification (retryable vs fatal) is provided.
Tool descriptions do not mention that LIMEAI_ACCESS_TOKEN environment variable must be set. Agents will fail with a cryptic exception instead of a clear error message. Prerequisite dependency should be documented.