NewsServer has significant definition quality gaps. Tool names are descriptive (search_google_news, analyze_sentiment, send_email_with_attachment), but descriptions are verbose and include implementation details rather than LLM-optimized guidance. Parameter schemas are present with type definitions, but descriptions lack clarity on constraints, formats, and error recovery. No output schemas documented. Error handling is minimal, tools return status strings without actionable guidance for LLM recovery. The analyze_sentiment and send_email_with_attachment tools exhibit poor separation of concerns (sentiment analysis + file writing; email sending + attachment handling). No pagination, no structured output, no tool annotations.

• Rewrite tool descriptions to 50 - 150 chars, focusing on WHAT and WHEN. E.g., 'Search Google News by keyword, return top 5 articles with title, snippet, link.' Move parameter details to parameter descriptions.
• Add explicit output schemas. Document that search_google_news returns {articles: [{title, desc, url}]}, analyze_sentiment returns {file_path, status}, send_email_with_attachment returns {status, message}.
• Add format constraints to parameters: 'keyword (1 - 100 chars, alphanumeric + spaces)', 'filename (alphanumeric + underscore, .md extension)', 'to (valid email address)'.
• Implement error categorization: return {error: string, category: 'retryable'|'user_fixable'|'fatal', suggestion: string}. E.g., 'Missing SERPER_API_KEY. Set SERPER_API_KEY in .env and restart server.'
• Split analyze_sentiment into two tools: analyze_sentiment (returns analysis text) and save_sentiment_report (takes analysis, saves to file). Separate concerns enable agent composition.