Single tool 'turn.wait' has a well-structured schema with proper types, constraints (min/max for timeoutSeconds, maxLength/maxItems for options), and comprehensive parameter descriptions in both English and Chinese. Tool name is clear and verb-based. Description explains purpose and use cases. However, output schema is not documented, the response structure is not visible in the tool definition. Error handling and recovery guidance are absent. The tool is READ_ONLY (safe), but no explicit idempotent or readonly annotations are present in the schema.

• Document the output schema explicitly. Specify the response structure: e.g., { success: boolean, userInput: string, timedOut: boolean, selectedOption?: string }. Include field descriptions so LLMs know what to extract.
• Add error handling guidance to the tool description. Explain: 'On timeout, returns timedOut=true. On invalid input, returns error with constraint details. Retryable on network failure.'
• Add tool annotations to the schema: set readOnlyHint=true and idempotentHint=true to signal safety to agents.
• Consider expanding the server with complementary tools: e.g., 'turn.poll' for non-blocking status checks, 'turn.notify' for async notifications. This enables richer human-in-the-loop workflows.
• Validate input constraints at runtime and return actionable error messages. E.g., if timeoutSeconds < 10, return 'Invalid timeout: must be 10 - 3600 seconds' instead of a generic error.

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

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained