Single tool 'get_question' has critical definition gaps. Description is non-English ('英作文の問題を取得する') and only 12 characters when translated ('Get essay questions'), below the 20-char minimum. No parameter descriptions exist. Output is unstructured free-text concatenation rather than typed fields. Tool lacks error handling guidance, recovery paths, and actionable error messages. No pagination despite returning database results. No idempotency guarantees despite modifying state (solved_count increment). Naming is acceptable (verb_noun form), but everything else fails production standards.

• Add English description (80-150 chars): 'Retrieve a random English essay question from the database, ordered by least-solved first. Returns question ID, Japanese translation, and model answer. Increments solve counter on each call.'
• Define input schema with optional parameters: level (enum: beginner|intermediate|advanced), keyword (string, max 50 chars). Add description to each: 'level: Difficulty tier. keyword: Grammar/vocabulary focus (e.g., "that", "passive voice").'
• Return structured TextContent with JSON body: {"id": int, "translation": string, "sentence": string, "keyword": string, "level": string}. Document this in tool description.
• Add error handling: if no questions match, return 'No questions found for level={level}, keyword={keyword}. Try broadening filters or call list_available_keywords() first.'
• Add idempotency note to description: 'Warning: This tool increments a counter on each call. Retries will increment multiple times. Consider caching results if retry safety is needed.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry