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, explaining WHAT and WHEN. All parameters have type and description. However, output schemas are undocumented, LLMs cannot predict response structure. Parameter constraints (e.g., limit 1 - 50) are stated in descriptions but not enforced via JSON Schema enums/min/max. No error recovery guidance. Missing required field declarations in some schemas (e.g., search_datasets has no 'required' array). No tool annotations (readOnlyHint, idempotentHint) despite all being safe, read-only operations.

• Document output schemas for all three tools. Add a 'returns' section to each tool definition showing field names, types, and nesting. Example: search_datasets returns {total: number | null, count: number, datasets: [{id: string, name: string, summary: string, org: string, org_id: string, record_count: number, type: string, url: string}]}.
• Add JSON Schema constraints: search_datasets.limit with minimum: 1, maximum: 50; query_layer.limit with minimum: 1, maximum: 2000; query_layer.offset with minimum: 0. Use enum for out_fields if only specific field sets are valid.
• Enhance error messages with recovery guidance. Example: if query_layer returns 400, check if the where clause is valid SQL syntax. If 404, verify the Feature Service url is correct and still accessible.
• Mark search_datasets.query as required: ['query'] in inputSchema, or document the default behavior when query is omitted (e.g., returns all datasets).
• Add tool annotations to inputSchema: readOnlyHint: true and idempotentHint: true for all three tools to signal to agents that these are safe, repeatable operations.