Single tool 'mikusearch' has critical definition gaps. Tool name lacks action verb prefix (should be 'search_miku' or similar). Description is in Chinese and exceeds 1024 chars when including parameter details inline. Input schema is severely incomplete: only 'query' parameter is defined with type string, but docstring mentions 'search_mode' and 'language_mode' parameters that are NOT in the schema. These undeclared parameters are hardcoded in the implementation (simple_search, zh_CN) with no way for the LLM to control them. No output schema documented. No error handling guidance. API key exposed via environment variable (correct) but no validation or error recovery documented.

• Rename tool to 'search_miku' or 'search_ai_miku' to start with action verb and clarify intent.
• Add 'search_mode' and 'language_mode' to input schema as enum parameters with descriptions: search_mode: enum [simple_search, deep_search] (default simple_search); language_mode: enum [zh_CN, zh_HK, en_us, ja, ko, ru, ar] (default zh_CN).
• Document output schema explicitly. Inspect actual API response and define expected fields (e.g., results: array of {title, url, snippet}, total_count: int).
• Translate description to English and move parameter details into individual parameter descriptions in the schema, not the tool docstring.
• Add error handling: catch aiohttp exceptions and return structured errors like {error: 'API_UNAVAILABLE', message: 'MIKU API is unreachable. Check MIKU_API_KEY is set and valid.', retryable: true}.