Five tools with consistent naming (verb_noun pattern: ndl_search_*) and adequate descriptions (120-180 chars each). All tools have identical input schemas with query (string) and max_records (integer, 1-100). Schemas are present and properly typed. However, parameter descriptions lack detail on format/constraints, output schemas are undocumented, and error handling guidance is absent. No tool annotations (readOnlyHint, etc.). Descriptions meet minimum length but lack LLM-optimization detail on WHEN to use each tool vs. alternatives.

• Document output schema for each tool. Specify: result structure (e.g. array of records), per-record fields (title, author, date, url, provider), pagination (total_count, next_cursor), and data types. Example: 'Returns array of bibliographic records, each with {id, title, author, date, url, provider_name}.'
• Enhance parameter descriptions with constraints. For 'query': state length limits, allowed characters, syntax rules (e.g. 'CQL syntax supported'). For 'max_records': clarify minimum (1), default (10), maximum (100), and behavior if exceeded.
• Add tool annotations via readOnlyHint=true to all five tools. This signals to agents that calls are safe to retry and have no side effects.
• Clarify tool selection in descriptions. Rewrite each tool's description to explain its scope: 'Search the Japanese National Bibliography (iss-ndl-opac-national), use this for comprehensive Japanese publication records. For holdings in NDL's physical collection, use ndl_search_holdings instead.'
• Add error handling guidance. Document common failures: 'If query returns 0 results, try broadening the search term or using ndl_search_all to check other providers. If rate-limited (429), wait 10+ seconds before retrying. If timeout, reduce max_records.'

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

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry