Server has 14 tools with basic schemas and descriptions, but significant gaps limit production readiness. Most tools have minimal descriptions (10-50 chars, well below the 194-char baseline). Parameter descriptions are sparse or missing entirely. No visible error handling guidance, output schemas, or pagination support. Tools like 'config' and 'instructions' lack actionable descriptions. Critical write tools (execute_nonquery, execute_routine, add_row) have no confirmation/dry-run patterns. Schema completeness varies: some tools show parameter types, but many lack descriptions for what those parameters control. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications. Composition is reasonable (14 focused tools), but descriptions do not guide LLM selection or explain when to use similar tools (e.g., execute_reader vs execute_nonquery).

• Expand tool descriptions to 100-200 chars. For execute_reader: 'Executes one or more SQL SELECT statements and returns results as structured rows. Use this to query data. Returns column names and values for each row. Supports pagination via LIMIT/OFFSET.'
• Add parameter descriptions for 'parameters' objects. E.g., for add_row: 'Dictionary mapping column names (string keys) to values. Values must match column types (string, number, boolean, null). Required columns must be provided.'
• Document output schemas for all tools. E.g., get_tables returns [{name: string, schema: string, row_count: number}, ...]. Include field types and meanings.
• Add tool annotations: mark READ_ONLY tools with readOnlyHint=true; mark WRITE tools with destructiveHint=true and idempotentHint=false (unless confirmed idempotent).
• Implement dry-run support for write tools. Add optional 'dry_run: boolean' parameter to execute_nonquery and execute_routine. Return 'Would execute X rows affected' without committing.