Server has 4 tools with complete input schemas and descriptions in Chinese. All tools are READ_ONLY and follow verb_noun naming (list_*, search_*, get_*). Schemas include type definitions and parameter descriptions. However, descriptions are brief (10-50 chars) and lack actionable context for LLM selection. Output schemas are not documented, responses are JSON strings without declared field structure. No error handling guidance, no parameter constraints (enums), and no indication of when to use each tool vs. alternatives. Parameter descriptions exist but are minimal. Missing: output schema documentation, error recovery guidance, and richer descriptions explaining prerequisites and use cases.

• Document output schemas for all tools. Example for search_knowledge_base: 'Returns {chunks: [{content, doc_name, page, score, source}], graph: {entities: [{name, type, verified}], relations: [{source, target, type}]}, permission_scope}'.
• Expand descriptions to 80-150 chars with use-case guidance. E.g., search_knowledge_base: 'Search knowledge base using vector + graph retrieval. Use this to find relevant documents and related entities. Returns ranked chunks with sources and entity relationships.'
• Add error handling guidance in descriptions. E.g., 'If no results found, try: (1) simplify query, (2) check knowledge base permissions, (3) verify entity exists in graph'.
• Declare parameter constraints explicitly in schema. Add 'minimum', 'maximum', and 'enum' fields to top_k and graph_depth. Update descriptions: 'top_k: number of results (1-50, default 8)'.
• Add pagination metadata to list_documents response. Return {documents: [...], total_count, limit, offset} so LLM knows if results are truncated and can request more.