MCP server for analyzing large fund flows in A-share stocks using real-time market data
Single tool with mixed quality. The tool has a detailed, multi-paragraph Chinese description that explains the analysis criteria and thresholds well. All 9 parameters have type definitions and descriptions with defaults. However, the schema lacks critical elements: (1) No explicit output schema documentation is visible, the return type is `dict` with a vague 'Returns: dict' docstring that does not specify what fields the result contains, what the 'data' array structure is, or what type of objects are returned; (2) Parameter descriptions, while present, are written in Chinese and lack English clarity for non-Chinese-speaking LLMs; (3) The tool name 'analyze_large_fund_flow' is sufficiently specific but describes analysis, not action (get_, create_, update_, etc.), which is acceptable for analytical tools but slightly weaker than action verbs; (4) No error handling guidance is present, the code returns error dicts like {'message': 'X', 'data': []} but the docstring does not explain when these occur or how to recover; (5) The tool lacks any indication of whether it is read-only, idempotent, or has side effects, only the summary states 'Risk: READ_ONLY' but this is not embedded in the tool metadata or description.
分析大额资金流入/流出的股票 该工具基于以下标准分析大额资金流动情况: 1. 主力资金净流入/流出 ≥ 指定阈值(默认5000万元) 2. 交易量占比(成交额/总市值)≥ 指定阈值(默认6%) 3. 涨跌幅绝对值 ≥ 指定阈值(默认3%) 4. 主力资金占成交额比例 ≥ 指定阈值(默认10%) 5. 机构/股东持股变化(辅助确认资金意图)
Output schema is undocumented. The tool returns `dict` but the docstring does not specify the structure of the result, the fields in the response object, the structure of the 'data' array, or the type of objects within it. LLMs cannot plan downstream tool calls or extract the right data without knowing what fields to expect.
Parameter descriptions are in Chinese. While technically present, they are not accessible to English-language LLMs that do not have Chinese language understanding. This violates the principle of self-documenting tools for multi-language agent use.
No error handling guidance. The code returns error responses like {'message': 'unretrieved data', 'data': []} but the docstring does not explain when errors occur, what they mean, or how the LLM should recover. For example, the tool may fail if AKShare API is unreachable or if a stock_type value is invalid, but the LLM has no guidance on this.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 44 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Tool name lacks action verb. 'analyze_large_fund_flow' describes the operation but does not start with a clear action verb like 'get_', 'search_', or 'find_'. This makes it slightly less discoverable in a list of tools, though the name is still clear enough.
Missing tool annotation metadata. The tool is read-only and idempotent (as stated in the tool summary), but these properties are not encoded as tool annotations (readOnlyHint, idempotentHint) that the LLM and client can discover. This reduces transparency about the tool's safety properties.
Parameter names and descriptions use Chinese terminology inconsistently. Some parameters like 'stock_type' have English names with English descriptions, while the description text is in Chinese. This mixing creates confusion for non-Chinese-speaking agents and humans.