The server implements 4 tools for Baidu search and web content fetching. Tool naming follows verb_noun convention (baidu_search, fetch_page_content), which is good. However, descriptions are uniformly terse (14-41 characters), well below the 50-200 char baseline for LLM-optimized descriptions. Parameter descriptions exist but are minimal. Input schemas are visible and properly typed (string, integer), but lack detail on constraints, ranges, valid values, and dependencies. No output schemas are documented in the code. Error handling is present at the code level (try/except, logging) but does not provide actionable recovery guidance to the LLM. The code shows rate limiting and graceful fallbacks (curl_cffi, Playwright), which is production-aware, but these implementation details are not reflected in the tool interface definitions visible to the agent.

• Expand tool descriptions to 50-200 characters. Example: 'baidu_search' should be: 'Search the Baidu search engine and return ranked results with titles, URLs, and snippets. Use query parameter for keywords, pn for pagination (1-indexed). Returns up to 20 results per page.'
• Document output schema for each tool. Specify return fields (e.g., title: string, link: string, snippet: string, position: integer for baidu_search) and include total_count, next_page_token, or pagination metadata.
• Add parameter constraints to descriptions: 'pn (page number): integer >= 1, default 1. Pagination limit: 20 results per page. Max 50 pages.' Enforce these in code and return validation errors.
• Implement actionable error messages: instead of returning [] on failure, return {'error': 'Network timeout fetching Baidu. Retry after 5 seconds.', 'retryable': true}. Document error response schema.
• Add a note to baidu_search description stating result limits: 'Results are capped at 20 per page. Use pn parameter to paginate. For comprehensive searches, plan to call multiple times with increasing pn values.'
• Document rate limiting transparently: add to all tool descriptions: 'Rate limited to 30 requests per minute across all Baidu tools. Plan multi-tool workflows accordingly.'
• Add parameter descriptions for all query parameters: 'query: Search keywords (string, required). LLM should sanitize input to prevent unexpected results.'

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

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