Three well-named, read-only tools with clear descriptions and complete input schemas. Naming follows verb_noun pattern (search_datasets, query_layer, layer_info). Descriptions are 150 - 250 chars, within the 10 - 1024 baseline. All parameters have type definitions and descriptions. However, output schemas are not documented, LLMs cannot predict response structure for downstream planning. Error handling is minimal (generic HTTP status errors with no recovery guidance). No parameter validation constraints (e.g., limit ranges are stated in descriptions but not enforced via enum/pattern). Missing tool annotations (readOnlyHint) despite all tools being read-only.

• Document output schemas for all three tools. Example for search_datasets: 'Returns {total: number, count: number, datasets: [{id, name, summary, org, org_id, record_count, type, url}]}'. This enables LLMs to chain tools and extract IDs for downstream calls.
• Enhance error messages with recovery guidance. Replace 'ArcGIS Hub: 404' with 'Dataset not found. Try search_datasets() with a different keyword or check org_id.' Categorize errors as retryable (network timeout) vs user-fixable (invalid URL format).
• Add tool annotations to schema. Set readOnlyHint: true for all three tools so clients know they are safe to cache and retry without side effects.
• Add minValue/maxValue constraints to numeric parameters in inputSchema. E.g., limit: {type: 'number', minimum: 1, maximum: 50, default: 20}. This enables schema-level validation and prevents LLM hallucination of out-of-range values.
• Document query_layer result limits and pagination. Add to description: 'Returns up to 2000 features per call. Use offset for pagination. Large result sets may exceed context limits, consider filtering with where clause.' Return a next_offset hint if more results exist.