A remote MCP server providing calculation tools without authentication
This calculator server has severe definition quality gaps. Both tools have minimal descriptions (under 20 characters), lack parameter descriptions entirely, and provide no output schema documentation. Tool names are generic verbs without clear action semantics. Parameter types are present in the Zod schema but not documented in natural language, forcing LLMs to infer meaning. The 'calculate' tool uses an enum correctly, but this alone cannot offset the missing descriptions and documentation across the board.
Simple addition tool
Calculator tool with multiple operations
All tool descriptions are under 20 characters ('Simple addition tool', 'Calculator tool with multiple operations'). Per hard scoring rule, description scores capped at 0-20. These descriptions provide no context for LLM tool selection, do not explain WHEN to use the tool, and do not document return types.
Zero parameter descriptions for any input. The 'a' and 'b' parameters in both tools have null descriptions in the schema. An LLM cannot infer that 'a' and 'b' are operands, what their valid range is, or what units they represent. The 'operation' enum in 'calculate' also lacks a description explaining which operation performs which arithmetic.
No output schema documented. Neither tool specifies what structure is returned. The code returns {content: [{type: 'text', text: String(result)}]} but this is not declared in any schema visible to the LLM. Without output documentation, the agent cannot plan downstream tool calls or validate results.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 5 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 26 | - | v1 |
Tool names lack action verb clarity. 'add' and 'calculate' are generic. The naming pattern 'verb_noun' (e.g. 'compute_sum', 'compute_arithmetic') is not followed. 'calculate' is particularly ambiguous, does it perform a single operation, multiple operations, or something else? LLMs must read the full description to disambiguate, but descriptions are missing.
Error handling in 'calculate' returns a plain text error for division by zero ('Error: Cannot divide by zero') but no guidance on recovery. Per pattern:recovery-guide, error responses should tell the LLM what to do next. Should it retry with a different divisor? Should it ask the user? The error is unstructured.
No error classification. The division-by-zero case returns an error, but there is no indication whether this is retryable, user-fixable, or fatal. An LLM cannot determine the appropriate recovery strategy without explicit classification.