Six tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error handling. All tools have names starting with action verbs (add_, list_, complete_, delete_, get_) which is good. Descriptions are present but brief (10-50 chars), below the 50-200 char baseline for LLM-optimized descriptions. Input schemas visible for all tools with type definitions, but parameter descriptions are minimal. Output schemas are not documented, LLMs cannot predict return structure. No error handling guidance, no recovery paths, no validation messaging. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Composition is reasonable (single responsibility per tool), but chaining support is weak, list_habits returns rows but no clear field names for downstream use.

• Document output schemas for all tools. E.g., list_habits should return {habits: [{id: int, name: string, description: string, created_at: string, is_active: bool}]}. Use structured JSON responses instead of raw tuples.
• Expand tool descriptions to 50-200 chars. Example: 'add_habit: Create a new habit to track. Returns the habit ID for use with complete_habit and delete_habit.' Include WHEN to use and what to do next.
• Add parameter descriptions with format/range guidance. E.g., habit_id: 'The numeric ID of the habit (from list_habits result). Must be a positive integer.'
• Implement error handling with recovery paths. E.g., delete_habit should return 'Habit 123 not found. Call list_habits() to see available habits.' complete_habit should validate habit_id before inserting.
• Add tool annotations: mark delete_habit with destructiveHint=true, list_* with readOnlyHint=true, get_current_streak with idempotentHint=true. This guides agent planning and retry logic.