Four tools with complete Zod schemas and descriptions. All tools are READ_ONLY and well-named with action verbs (list_, query_, inspect_). Descriptions are clear and contextual (avg ~150 chars). Parameters are typed and described. However, output schemas are not documented, LLMs cannot predict response structure. Error handling is absent; no recovery guidance. Missing tool annotations (readOnlyHint). query_search_analytics has complex optional parameters that could benefit from clearer dependency documentation.

• Document output schemas for all tools. E.g., list_sites returns {siteUrl: string, permissionLevel: string}[]. query_search_analytics returns {rows: {keys: string[], clicks: number, impressions: number, ctr: number, position: number}[], responseAggregationType: string}. This enables LLMs to chain calls and extract the right fields.
• Add error handling with recovery guidance. Wrap tool implementations in try-catch; return structured errors like {error: 'Invalid siteUrl', suggestion: 'Call list_sites() first to discover valid siteUrl values'}.
• Add readOnlyHint annotation to all four tools via server.tool(..., {readOnlyHint: true}) to signal safety to LLMs and enable aggressive caching.
• Document parameter dependencies in query_search_analytics: 'If filterDimension is set, filterExpression is required. filterOperator defaults to equals if omitted.'
• Document pagination behavior: 'Results are capped at rowLimit (max 25000). If more rows exist, increment startRow and call again. The response includes rowCount to detect when all rows are fetched.'

Inferred effective spec: <=2025-11-25.

• Relies on Dynamic Client Registration (deprecated) - use Client ID Metadata Documents (CIMD)