Server provides 5 database tools with reasonable naming (verb_noun pattern) and descriptions in Chinese. All tools have input schemas with type definitions and parameter descriptions. However, descriptions lack English translations (barrier for international LLM use), output schemas are not documented, and error handling guidance is minimal. Parameter constraints are present but could be more explicit. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Average tool quality is fair but not production-grade.

• Translate all tool descriptions and parameter descriptions to English. Use clear, concise language (50 - 200 chars per description) optimized for LLM comprehension. Example: 'execute_query: Execute SQL statements. Supports SELECT, SHOW, DESCRIBE, EXPLAIN by default; DML/DDL if sqlAccess level permits. Results truncated at maxRows; returns row count and truncation flag.'
• Document output schemas for each tool. Specify field names, types, and whether fields are always present or conditional. Example for execute_query: '{rows: [{col1: string, col2: number, ...}], rowCount: number, truncated: boolean, executionTimeMs: number}'
• Add tool annotations to execute_query: set readOnlyHint=false and destructiveHint=true when sqlAccess is 'dml'|'ddl'|'full'. This signals to clients that the tool can modify data.
• Enhance error messages with recovery guidance. Example: 'Schema "unknown_schema" not found. Available schemas: [list]. Call list_schemas() to explore all accessible schemas.' This enables LLM self-correction.
• Add offset/limit pagination to execute_query. Replace maxRows with limit (1 - 1000, default 100) and offset (0+). Return total_available_rows so LLM knows if more results exist. This prevents context window exhaustion.

Inferred effective spec: 2025-06-18+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry