Single tool 'greet' has excellent schema with minLength/maxLength constraints and comprehensive tool annotations (readOnlyHint, idempotentHint). Description is detailed and explains the validation mechanism. However, the server is minimal (1 tool) and lacks the breadth expected of production systems. Parameter description is clear but could be more concise. Tool follows verb_noun naming convention. No critical gaps, but limited scope prevents higher score.
Greet someone by name. The Annotated[..., Field(...)] form propagates the constraint into FastMCP's generated JSON schema, so empty strings and oversized inputs are rejected by the protocol layer before the handler runs. The TS sibling enforces the same shape via Zod.
Single tool limits evaluation scope. Production servers typically expose 5-20 tools. This starter template is intentionally minimal, but scoring reflects that limitation.
Tool description is 280+ characters, exceeding the 200-char sweet spot for LLM token efficiency. Trim to core action + when to use.
No output schema documented. Tool returns a string, but LLMs benefit from explicit return type documentation in the tool definition.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | <=2025-11-25 | v2 |