GenPark analytics search keyword metrics and clicks monitoring skill.
This MCP server has a single tool with significant definition quality gaps. The tool is named correctly with a verb (aggregate_metrics), but the description is extremely brief (95 chars, well below the 194-char baseline for A+ tools). The input schema is visible and properly structured with types, but critically lacks parameter descriptions, the 'search_logs' parameter has a description, but the nested object structure (query string, results_count integer) is not formally validated or constrained in the schema itself. The output schema is not documented, LLMs have no contract for what fields to expect beyond reading the implementation. No error handling guidance is present; the tool could fail silently on malformed input. The tool's composition is sound (single responsibility), but the overall packaging for LLM consumption is poor. The server appears to be a simple Python class with no MCP framework integration visible, tool registration, transport, and protocol adherence are completely absent from the source.
Aggregates search analytics metrics from a list of search log entries, returning total queries logged, count of zero-result queries, and top 5 searches by frequency
No output schema documented. LLMs cannot infer return type structure (total_queries_logged, zero_results_count, top_searches fields are invisible in schema).
Tool description is 95 characters, well below the 194-char baseline for A+ tools. Missing context on WHEN to use this tool and what problem it solves. Does not explain use case beyond the mechanic.
Nested input parameter schema lacks formal constraint definition. The description states 'each containing query (string) and results_count (integer) keys' in prose, but the JSON Schema does not enforce this structure, no additionalProperties: false, no required fields on nested objects, no type constraints on array items.
No error handling guidance. Tool will silently return incorrect results if input is malformed (missing 'query' or 'results_count' keys, non-integer counts, etc.). LLM has no recovery path.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | 2026-07-28+ | v2 |
No MCP framework detected. Source code shows raw Python class, not MCP server implementation. No tool registration, no transport protocol, no request/response handling visible.