Loading ToolBench
MCP server for interacting with MySQL databases with write operations support
Single tool 'mysql_query' has critical gaps. Description is present (95 chars) but lacks WHEN/WHY guidance for LLM selection. Input schema exists but contains only one parameter ('sql' as string) with minimal description ('The SQL query to execute'). No output schema documented. No parameter constraints (enums, patterns, ranges). No error handling guidance. No mention of idempotency, retryability, or side effects despite WRITE risk classification. Tool does too much (INSERT, UPDATE, DELETE, DDL, READ in one tool) violating single-responsibility principle. No pagination, no result limits documented. Lacks security context (permissions, audit trail, injection prevention).
Run SQL queries against MySQL database with support for: INSERT, UPDATE, DELETE, DDL and READ operations (Schema-specific permissions enabled)
Single parameter 'sql' lacks type safety, format constraints, and validation guidance. No enum, pattern, or length limits. LLM can pass arbitrary SQL including injection payloads.
No output schema documented. LLM cannot predict response structure, field names, or types. Impossible to chain downstream tools or extract specific data.
Tool description (95 chars) lacks WHEN/WHY context. Does not explain when to use vs alternatives, prerequisites, or consequences. Missing: 'This tool modifies state (INSERT/UPDATE/DELETE are irreversible)' and 'Requires schema-specific permissions.'
Tool combines 5 distinct operations (INSERT, UPDATE, DELETE, DDL, READ) in one tool. Violates single-responsibility principle. Should split into separate tools: execute_read_query, execute_write_query, or more granular variants.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | <=2025-11-25 | v2 |
No error handling guidance. No recovery hints for common failures (syntax error, permission denied, connection timeout, table not found). LLM receives raw errors with no actionable next steps.