Loading ToolBench
Tai Yi Shen Shu (太乙神數) MCP server — Tai Yi calculations over stdio.
Single tool 'get_taiyi' has a complete input schema with proper types and enum constraints, but the description is domain-specific Korean text that provides minimal guidance for LLM tool selection. Output schema is not documented, the tool returns a list containing text and PNG image, but no structured schema is declared. Parameter descriptions are present but terse. No error handling guidance is provided to the LLM. The tool performs a single, well-defined action (divination calculation), but lacks the clarity and completeness expected for production agent use.
태을신수 포국/산법을 계산한다.
Output schema not documented. Tool returns [text, Image] but LLM has no declared schema for what fields/structure to expect. Prevents downstream tool chaining and forces LLM to infer structure.
Tool description is in Korean ('태을신수 포국/산법을 계산한다.') with no English translation. LLMs trained primarily on English will struggle to understand when/why to invoke this tool. Descriptions should be in the LLM's primary language.
No error handling guidance. If datetime_str is malformed or mode is invalid, the tool raises ValueError. LLM receives no recovery hint (e.g., 'Try format as YYYY-MM-DD HH:MM'). Errors should guide the agent to self-correct.
Parameter 'datetime_str' description lacks format validation rules in plain text. While the docstring mentions 'YYYY-MM-DD HH:MM', the schema description should explicitly state the required format and consequences of mismatch.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Mode parameter 'ming' has a caveat (approximation via year-based calculation) documented only in the docstring, not in the enum description or return value. LLM may not recognize the limitation and misrepresent results to users.