Single tool 'search_gpters' has a complete input schema with 5 parameters, all typed and described. However, the tool description is in Korean and lacks LLM-optimized English guidance on when/why to use it. Output is returned as unstructured plain text (formatted string) rather than structured JSON, forcing LLMs to parse results manually. No error handling guidance, errors return generic messages like 'search 중 오류가 발생했습니다' without recovery hints. Parameter descriptions are minimal (e.g., 'top_k': '반환할 결과 수 (기본값: 5)') and lack format/range constraints. No output schema documented. Tool follows verb_noun naming (search_gpters) correctly but lacks English descriptions critical for LLM selection.

• Rewrite tool description in English: 'Search the GPTers AI study community knowledge base. Returns ranked results with title, author, board, tags, creation date, and relevance score. Use this to find answers, best practices, and discussions from the community. Supports filtering by board name, author, and recency.'
• Return structured JSON output instead of formatted text: {"results": [{"title": string, "text": string, "owner_name": string, "space_name": string, "tags": array, "created_at": ISO8601, "url": string, "score": number, "original_score": number, "recency_weight": number}], "total": number}
• Add actionable error messages: 'Search failed: API returned 401 Unauthorized. Token may have expired. Retrying with fresh authentication.' or 'Search failed: Query too broad. Try adding more specific keywords or filtering by board/author.'
• Document parameter constraints: 'top_k (integer, 1 - 100, default 5): Maximum number of results to return.' and 'created_within_days (integer, 1 - 365, optional): Limit results to posts created within N days.'
• Add output schema to docstring: 'Returns: {results: [{title, text, owner_name, space_name, tags, created_at (ISO8601), url, score (0 - 1), original_score (0 - 1), recency_weight (0 - 1)}], total: number}'

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

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