Single tool with adequate naming (research_generate_report) and comprehensive description (194 chars, within baseline 34-392 range). Input schema present with proper types and descriptions for both parameters. However, output schema is completely undocumented, the tool returns markdown-formatted report but LLMs have no structured schema to parse response fields. No error handling guidance, no parameter constraints (depth enum should be enforced), and no documentation of what fields the markdown contains or how to extract data from it. Tool description includes helpful anti-patterns (what NOT to use it for) but lacks actionable error recovery guidance.

• Document output schema: {type: 'object', properties: {report: {type: 'string', description: 'Markdown-formatted report with sections: Summary, Key Findings, Sources'}, source_count: {type: 'number'}, generated_at: {type: 'string', format: 'date-time'}}}
• Convert depth parameter to enum: {type: 'string', enum: ['quick', 'standard', 'deep'], description: 'Research depth: quick (3 sources), standard (5 sources), deep (8 sources). Default: standard'}
• Add error handling to POST /api/report: return {error: string, recovery_hint: string} with actionable guidance (e.g., 'TAVILY_API_KEY not configured. Ensure environment variable is set.' or 'Topic too broad. Try a more specific query.')
• Document response structure in tool description: 'Returns markdown report containing: title, generation timestamp, source count, summary section, key findings with excerpts and citations, and sources list.'
• Add topic validation: reject empty/whitespace-only topics with clear error message before calling Tavily API

Inferred effective spec: 2025-06-18+.

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