Three tools with significant definition gaps. All tools have descriptions (positive), but schemas are incomplete or inferred from code rather than explicitly registered. The 'debate' tool has a partial schema visible in code; 'reframe' and 'diverge' schemas are not visible in the provided source. Parameter descriptions exist but lack detail on constraints, ranges, and error recovery. No evidence of output schema documentation. Error handling is minimal, no recovery guidance or actionable error messages. Tool composition is reasonable (each does one thing), but parameter validation and LLM-friendly descriptions need work.

• Explicitly register all three tools with complete JSON Schema input definitions. Include type, description, and constraints for every parameter. Verify schemas are visible in index.js tool registration.
• Document output schemas for each tool. Specify field names, types, and structure. E.g., debate returns {rounds: [{skeptic: string, steelman: string, cost: number}], totalCost: number}.
• Add parameter constraints: maxRounds (integer, 1-10, default 3), domain (string, 1-100 chars, optional), constraints (string, 1-500 chars, optional). Use descriptions to state these explicitly.
• Implement error recovery guidance. When API calls timeout or fail, return structured errors: {error: 'timeout', message: 'Debate round timed out after 90s. Retry with fewer rounds or longer timeout.', retryable: true}.
• Enhance tool descriptions with WHEN to use guidance. E.g., 'debate: Use when you need adversarial critique of a plan. Calls GPT and Gemini in parallel, returns skeptic and steelman analyses with cost tracking. Typical cost: $0.50 - $2.00 per debate.'