Two tools with significant definition gaps. bocha_search has a complete schema with proper types and descriptions, but web_search_tools has a critical flaw: the search_engines parameter lacks an explicit type definition in the schema (only enum values are shown in the description). Both tools have descriptions, but web_search_tools' description is bilingual and verbose (200+ chars), making it harder for LLMs to parse intent. Neither tool documents output schemas. Error handling is minimal, both return empty lists or error dicts without recovery guidance. No tool annotations (readOnlyHint, idempotentHint) despite being read-only operations.

• Add explicit 'type': 'array' to search_engines parameter schema and move enum constraint into JSON Schema, not just description text.
• Document output schemas for both tools. Example for bocha_search: {type: 'array', items: {type: 'object', properties: {title: {type: 'string'}, href: {type: 'string'}, body: {type: 'string'}}}}.
• Remove 'api_key' parameter from bocha_search. Load BOCHA_API_KEY from environment inside the tool; do not expose it as a parameter.
• Enhance error messages with recovery hints: 'Bocha API error: check BOCHA_API_KEY is set and valid' instead of bare 'HTTP error occurred'.
• Simplify web_search_tools description to English only, under 150 chars. Example: 'Aggregates search results from multiple engines (Baidu, DuckDuckGo, Bocha, Tavily). Specify 2 - 5 engines and a query; returns ranked results with title, URL, and snippet.'

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

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