The server registers 21 tools with complete input schemas and reasonable descriptions. All tools have descriptions (10-80 chars), and most parameters are typed with descriptions. However, descriptions are uniformly terse (under 50 chars for most tools), missing context about WHEN to use a tool vs. similar ones, and omitting prerequisites or dependencies. Output schemas are completely undocumented, no specification of what fields are returned or their types. Error handling and recovery guidance are absent from descriptions. The naming is clear and verb-driven (get_, create_, update_, delete_), but composition issues exist: multiple tools operate on overlapping concerns without clear disambiguation in their descriptions. Security patterns are partially addressed (Bearer token validation in middleware) but not reflected in tool descriptions. Overall, this is a competent but minimal implementation, suitable for a dashboard tool but lacking the polish expected for production agent integration.

• Add output schema documentation to every tool. Document what fields are returned, their types, and which IDs/references can be passed to downstream tools. Example: 'Returns {id: integer, name: string, collection_id: integer, items_count: integer}'.
• Expand descriptions to 80 - 200 chars. Explain WHEN to use the tool vs. alternatives. Example for get_card_query_results: 'Execute a saved question and retrieve results. Use after get_card() to fetch fresh data with current filters and parameters. Returns paginated result with {id, name, data, column_names}.'
• Add pagination parameters (limit, offset or cursor) to list/get_all tools. Document default and max limits (e.g., 'Default limit=20, max=100'). Cap results and return a total_count and next_cursor for chaining.
• For delete operations, add a 'dry_run' parameter (boolean, default false) or describe a two-step confirmation flow: dry_run first, then confirm with an explicit flag. Document that deletions are irreversible.
• Add parameter validation descriptions. Example: 'Row (integer, 0 - 10): grid row position. Col (integer, 0 - 3): grid column position. Size_x and size_y must be 1 - 4.'
• Implement SQL query validation. Sanitize and parse create_card and update_card queries to detect injection patterns. Add description: 'Validates SQL syntax and forbids certain keywords (DROP, TRUNCATE).'
• Document chaining IDs. If get_collection returns a collection_id, add description: 'Returns {id, name, items}. Use the id in get_collection_items() or update_collection(collection_id=...).'

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

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