MemoryOS exposes 4 tools with basic descriptions and minimal input schemas. Tool names follow verb_noun convention (get_*, query_*, update_*), which is good. However, descriptions are generic and lack actionable context for LLM selection. Input schemas exist but lack type constraints, enums, and validation guidance. No output schemas are documented. Error handling is minimal, tools return plain strings without recovery guidance. The update_wiki tool has a critical path-traversal check but no dry-run or confirmation pattern for destructive operations. Overall, definitions are functional but fall short of production quality.

• Add explicit output schemas to all tools. Define return type as JSON object with typed fields (e.g., {context: string, token_count: number, status: 'ready'|'onboarding'}) instead of free-text markdown.
• Enhance parameter descriptions with format constraints and examples. E.g., 'query: Search string (1-500 chars, supports Chinese and English). Example: "Python项目经验"'.
• Add WHEN-to-use guidance to tool descriptions. E.g., 'Call get_user_context at conversation start to inject background. Call query_wiki only when you need specific historical facts not in the context.'
• Implement dry-run or confirmation for update_wiki. Return a preview of changes and require explicit confirmation before writing. Document idempotency: 'Calling update_wiki twice with identical content is safe.'
• Wrap error responses with actionable recovery hints. E.g., 'Wiki not found. Run: python main.py --max-files 100 to initialize, then retry.' Include error codes for LLM classification (retryable vs user-fixable).