Server has 4 tools with basic descriptions and minimal schemas. Tool naming follows verb_noun convention (save_memory, get_all_memories, search_memories, get_bytedance_info). However, descriptions lack depth and actionable guidance. Parameter descriptions are present but sparse. No output schemas documented. Error handling returns generic strings without recovery guidance. The in-memory storage implementation is simplistic and lacks production-grade features like persistence, indexing, or structured responses.

• Expand get_bytedance_info description to 100+ chars explaining what information is returned and when to call it. Current 25-char description violates the 20-char floor rule.
• Document output schemas for all tools. E.g., get_all_memories should declare: 'Returns array of {role: string, content: string}'. search_memories should declare result structure and total count.
• Add recovery guidance to error messages. E.g., 'Error saving memory: text too long (max 10000 chars). Try splitting into smaller chunks.' or 'Error retrieving memories: storage unavailable. Retry in 5 seconds.'
• Expand save_memory and get_all_memories descriptions to 150+ chars. Include WHEN to use (e.g., 'Call before making decisions to ensure you leverage existing knowledge') and any prerequisites.
• Add min/max bounds to search_memories limit parameter: 'Maximum number of results to return (1-100, default: 3)' to prevent LLM from passing invalid values.