Two tools with basic descriptions and parameter schemas visible in curl examples. Both tools have action-verb names (search_, get_) and documented parameters with types and descriptions. However, output schemas are not explicitly documented in the source code, only inferred from example responses in quickstart.mjs. Parameter descriptions are adequate but brief (avg ~60 chars). No error handling guidance, no tool annotations (readOnlyHint/destructiveHint), and no documented output structure. Tools are read-only (low risk) but lack production-grade polish expected for A/B grades.

• Document complete output schemas for both tools. Include field names, types, and descriptions. Example: search_chinese_company returns {companies: [{companyId: string, nameZh: string, nameEn: string, ...}], total: number}.
• Add error handling guidance to tool descriptions. E.g., 'If no companies match the query, returns empty list. Try broadening the search term or using pinyin transliteration.'
• Annotate tools with readOnlyHint: true in tool registration to signal to agents that these are safe, non-destructive operations.
• Clarify parameter relationships in get_company_snapshot: 'Either companyId or query is required. If both provided, companyId takes precedence. If query matches multiple companies, returns best match.'
• Add pagination support: document limit (default 20, max 100) and offset parameters for search_chinese_company, and return next_cursor or total_count for agent-driven pagination.

Inferred effective spec: 2025-06-18+.

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained