Single tool 'general_search' has a basic description (89 chars) and minimal parameter documentation. The tool name lacks clarity on what 'general' means vs domain-specific search. Input schema is present with type and description for 'keyword', but output schema is undocumented (returns 'dict' with no field specification). No error handling guidance, no pagination support despite search results likely being large, and no indication of result limits. The implementation directly wraps bilibili_api without sanitization or result shaping.

• Document the complete output schema: specify which fields are returned (e.g., video_id, title, author, duration, view_count, url) with types and descriptions.
• Add pagination parameters: 'limit' (1-50, default 20) and 'offset' (0+, default 0). Return 'total_count' and 'has_more' in response to enable multi-page queries.
• Rename tool to 'search_bilibili_videos' and update description to clarify: 'Search Bilibili for videos matching a keyword. Returns up to 20 results per page.'
• Implement error handling: catch network errors, rate limits, and invalid input; return structured errors like '{"error": "rate_limited", "retry_after_seconds": 60}' with recovery guidance.
• Filter API response to return only: video_id, title, author, duration_seconds, view_count, url. Strip internal metadata, timestamps, and audit fields to reduce token waste.