Single tool 'chat' has a basic description (72 chars, within range) but lacks critical depth. Input schema is present with two parameters (message, history) but descriptions are minimal and generic. No output schema documented. No error handling guidance, no parameter constraints (enums, ranges), no indication of idempotency or side effects. Tool does not follow verb_noun naming (should be 'send_message' or 'process_message'). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). The implementation shows hardcoded responses and state mutation (message_history) but the tool definition does not document these behaviors or their implications for agent planning.

• Rename 'chat' to 'send_message' or 'process_message' to follow verb_noun convention and signal action intent.
• Expand tool description to 100-150 chars: 'Send a message to the chatbot and receive a text response. Maintains conversation history (last 10 messages). Supports English and Korean. Returns text content only.'
• Add parameter constraints: message (required, 1-2000 chars, non-empty string); history (optional, array of {role, content} objects, max 10 items).
• Document output schema: {content: string, type: 'text'} with field descriptions. Specify that 'content' is plain text (no markdown or HTML).
• Add error handling examples: 'If message is empty, returns error code -32603 with message "Message cannot be empty". Retry with non-empty input.'

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

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