Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. list_dataville_sources has an empty input schema ({}), which is unusual and undocumented. search_dataville has typed parameters with descriptions, but the 'params' parameter uses a generic record type without guidance on valid keys or values per source. Neither tool documents output structure, making it difficult for LLMs to plan downstream actions. Error handling exists (try/catch in search_dataville) but error messages are generic. No tool annotations (readOnlyHint, idempotentHint) despite both being read-only operations.

• Add explicit input schema to list_dataville_sources (even if empty, document why). Consider whether this tool should be auto-called on startup or if search_dataville should embed source validation.
• Document output schema for search_dataville: specify fields returned (e.g., results[], total_count, next_cursor), types, and pagination behavior. Include example response structure.
• Replace generic 'params' Record with source-specific parameter documentation. Either: (a) enumerate valid params per source in the description, (b) add a separate 'get_source_schema' tool, or (c) return param hints in list_dataville_sources output.
• Add readOnlyHint: true to both tool definitions to signal idempotency and enable client optimizations.
• Enhance error messages with recovery guidance: 'Source "arxiv" not found. Call list_dataville_sources to see available sources.' Include the list of valid sources in 'Unsupported data source' errors.