Three well-named, read-only tools with clear descriptions and complete input schemas. All tools follow verb_noun naming (search_, query_, layer_info). Descriptions are concise (100 - 180 chars) and explain WHAT, WHEN, and WHAT-TO-DO-NEXT. Input schemas include types and descriptions for all parameters. However, output schemas are not documented, LLMs cannot predict response structure for downstream chaining. No error handling guidance (e.g., 'If query fails, try layer_info first'). Parameter constraints (e.g., limit 1 - 50) are stated in descriptions but not enforced as JSON Schema min/max. No tool annotations (readOnlyHint, idempotentHint) despite all three being read-only and idempotent.

• 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.
• Add error handling guidance to descriptions. E.g., 'If query fails with 404, the layer may not exist, call layer_info first to verify schema and capabilities.'
• Add JSON Schema minimum/maximum constraints to numeric parameters. E.g., limit: {type: 'number', minimum: 1, maximum: 50}. This prevents LLMs from passing invalid values.
• Declare tool annotations in the McpToolDefinition interface. Add readOnlyHint: true and idempotentHint: true to all three tools so agents know they are safe to retry.
• Validate the 'where' clause in query_layer to prevent SQL injection. Reject clauses containing DROP, DELETE, INSERT, or other DML keywords. Return a clear error: 'where clause must be a SELECT filter (no DML allowed)'.