An MCP server system that orchestrates multiple MCP tools (SQL conversion and vector search) through a Gemini-powered client. The system includes a FastAPI-based client that receives instructions, uses Gemini to determine which tool to call, and executes the appropriate MCP server tool via STDIO transport.
Two tools with minimal definitions. Both have basic input schemas (text/query strings) but lack parameter descriptions, output schema documentation, and error handling guidance. Tool names are verb-noun but descriptions are under 50 chars and generic. No enum constraints, no pagination support, no recovery guidance. No tool annotations. Simulated implementations suggest this is a proof-of-concept, not production-ready.
Convert a textual instruction to a simulated SQL query.
Simulate a vector database query.
No output schema documentation. Both tools return dicts but LLMs cannot infer structure (sql_query field vs results array). Rubric baseline: 100% of A+ tools document return types.
Parameter descriptions missing. 'text' and 'query' params have no descriptions explaining format, constraints, or examples. Rubric: 100% of A+ tool params have descriptions.
Tool descriptions too generic and short (both ~40 chars). Rubric baseline: avg 194 chars, p10=34, p90=392. Descriptions lack WHEN to use, WHAT it returns, and prerequisites.
No error handling or recovery guidance. Simulated implementations always succeed; no validation, no error messages, no guidance for LLM on what to do if input is invalid.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | <=2025-11-25 | v2 |
SQL tool vulnerable to injection. txt_to_sql directly interpolates user text into SQL string without escaping. Production code must sanitize against SQL injection.