Loading ToolBench
明日方舟PRTS.wiki智能助手 - MCP服务器。一个基于Model Context Protocol (MCP)的明日方舟资料查询服务器，为AI助手提供准确的游戏资料访问能力。
Doctah-MCP has 14 tools with basic schemas and descriptions, but significant quality gaps. Tool names lack consistent verb prefixes (search_operator_mcp vs list_operators_mcp inconsistency). Descriptions are present but generic (avg ~60 chars, below 194-char baseline). 4 tools (get_stage_info, get_event_info, search_items, get_daily_stages) are unimplemented stubs returning placeholders. Parameter descriptions are minimal; many lack format/constraint guidance. No output schemas documented. Error handling absent. No tool annotations (readOnlyHint, destructiveHint). Chinese-language descriptions reduce LLM clarity.
获取今日开放关卡（暂未实现，返回占位符）
获取活动信息（暂未实现，返回占位符）
获取关卡信息（暂未实现，返回占位符）
敌人多维筛选（参考 PRTS『敌人一览』）维度见网页：地位/种类/攻击方式/伤害类型 + 六维。
搜索相关敌人并返回敌人名称列表（模糊搜索）
干员多维筛选（参考 PRTS「干员一览」的筛选维度）
搜索相关干员并返回干员名称列表（模糊搜索）
4 tools are unimplemented stubs (get_stage_info, get_event_info, search_items, get_daily_stages) returning placeholder responses. These should either be removed or have descriptions stating 'Not yet implemented' with a clear roadmap.
No output schemas documented. LLMs cannot predict response structure, forcing them to parse unstructured text. All tools should declare return types (e.g., 'Returns: {operators: [{name, rarity, profession}], total_count}').
Tool names inconsistently suffixed with '_mcp'. Naming convention should be uniform (either all with suffix or none). Current mix (search_operator_mcp vs list_operators_mcp) creates confusion. Recommend removing '_mcp' suffix, it's implementation detail, not user-facing.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | <=2025-11-25 | v2 |
公招计算（严格）：必须命中所有词条（AND）。
公招计算（分组）：按"所选词条子集"分组展示可能出现的干员，类似网页的分块列表。
公招计算：给定词条（如 "资深干员, 术师, 群攻" 或 "先锋 费用回复"），返回可能出现的干员。
公招计算（建议组合）：自动搜索能产出结果的最佳子集组合，返回前 top_k 个组合。
搜索明日方舟敌人信息
搜索道具信息（暂未实现，返回占位符）
搜索明日方舟干员信息
Parameter descriptions lack format/constraint guidance. E.g., 'sections' param says '用逗号分隔' (comma-separated) but no example or enum. 'limit' has no min/max bounds. Descriptions should state: 'Comma-separated values (e.g., "天赋,技能"). Max 10 sections.'
No error handling or recovery guidance. Tools return raw strings with no structure for error cases. Should return JSON with {success, data, error, suggestion} so LLMs can retry intelligently or ask users for clarification.