MCP server for CockroachDB integration
Single tool 'query' has a clear verb-noun name and basic schema with types. However, description is minimal (18 chars, below 20-char threshold), lacks context on when/why to use it, and omits critical safety guidance for a database tool. Parameters have types but descriptions are terse (8-15 chars each). No output schema documented. No error handling guidance. No mention of SQL injection risks or query limits. Tool lacks idempotency hints and destructive operation warnings despite accepting arbitrary SQL.
makes a query to the database
Tool description is 18 characters ('makes a query to the database'), below 20-char minimum. Lacks context on when to use, prerequisites, or safety constraints. LLMs cannot determine if this is for read-only discovery or destructive operations.
Parameter descriptions are extremely terse (8-15 chars: 'The database to query', 'The SQL statement to execute'). No guidance on format, constraints, or examples. Missing critical safety info: SQL injection risks, query timeout limits, or result row caps.
No output schema documented. Tool returns structured JSON with rowCount, fields, rows, command, duration, but LLM has no advance knowledge of this structure. Cannot plan downstream operations or extract fields reliably.
No error handling guidance. Tool catches errors but returns generic error responses. LLM cannot distinguish retryable (timeout) from fatal (syntax error) failures. No recovery hints (e.g., 'Try EXPLAIN ANALYZE to debug').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
Tool accepts arbitrary SQL without validation, rate limiting, or result caps. No mention of max rows returned, query timeout, or SQL injection prevention. Dangerous for agentic use, agent could accidentally run expensive queries or malicious payloads.