All 4 tools have basic schemas and descriptions, but quality is significantly below production standards. Descriptions are in Indonesian and lack LLM-optimized clarity (10-30 chars each). Parameter descriptions are minimal. No output schemas documented. No error recovery guidance. No tool annotations. Naming follows verb_noun pattern but descriptions fail to explain WHEN to use each tool or what distinguishes them from similar operations.

• Translate all descriptions to English and expand to 50-150 chars. Include WHAT the tool does, WHEN to use it, and what it returns. E.g., 'Retrieve the most recent session checkpoint for a project, including summary, pending tasks, and modified files. Call this at session start to resume prior work.'
• Document output schemas for get_session_handover and query_knowledge. Specify field names, types, and whether results are paginated. E.g., 'Returns {id: number, project_path: string, summary: string, next_steps: string, files_modified: string, created_at: ISO8601}'.
• Expand parameter descriptions with format and constraints. E.g., 'project_path: Absolute or relative path to project root (e.g., /home/user/myproject or ./myproject). Must be a valid directory.'
• Add error recovery guidance. E.g., 'If no session found, call store_knowledge to initialize the knowledge base, then retry.' Return structured error objects: {error: string, code: string, retryable: boolean, suggestion: string}.
• Add tool annotations via mcp.WithReadOnly() and mcp.WithDestructive() to clarify which tools are safe to retry and which modify state. This enables LLM planning without parsing descriptions.