Server has 5 tools with basic schemas and descriptions. Tool naming is inconsistent (weibo_token, weibo_search, weibo_status, weibo_hot_search, weibo_crowd). The weibo_crowd tool is a mega-tool with 19 optional parameters handling 9 different actions, violates single-responsibility principle. Descriptions are present but generic (avg ~80 chars, below 194-char baseline). Input schemas use Zod validation but lack comprehensive parameter descriptions. No output schema documentation. Error handling is minimal (only basic required-param checks visible). Tool annotations present (readOnlyHint, destructiveHint) but inconsistently applied.

• Split weibo_crowd into 9 separate tools: get_crowd_topics, get_crowd_topic_details, get_crowd_timeline, post_crowd_status, post_crowd_comment, post_crowd_reply, get_crowd_comments, get_crowd_child_comments, refresh_crowd_token. Each tool has one action, clear params, and obvious intent.
• Document output schemas for all tools. E.g., weibo_search returns {results: [{id, text, author, timestamp}], total_count, next_cursor}. Include field types and meanings so LLM can chain tools and extract data.
• Add parameter dependency documentation to weibo_crowd (if split). E.g., 'action=timeline requires topic, page, count; action=post requires topic, status, tagId (optional), mediaId (optional).'
• Enhance error messages with recovery guidance. E.g., 'topic is required when action=timeline. Available topics: [list]. Call get_crowd_topics() to discover all topics.' instead of bare error.
• Expand tool descriptions to 100-200 chars with context. E.g., 'Search Weibo posts by keyword. Returns matching posts with author, timestamp, and engagement metrics. Use this to find public content; use get_user_status() to fetch your own posts.'