Single tool 'send_email' has a basic schema with typed parameters and descriptions, but critical gaps undermine production readiness. The tool description (89 chars) is below the 194-char baseline and lacks WHEN/WHY guidance. Parameter descriptions are present but generic ('The recipient email address' vs actionable format/validation rules). No output schema documented. No error handling guidance, LLM cannot distinguish retryable failures from fatal ones. No idempotency or confirmation pattern for this destructive operation. The server itself is not MCP-compliant (custom TCP protocol, not MCP).

• Expand tool description to 150-250 chars: 'Send an email via SMTP to a recipient. Use this to notify users, send confirmations, or deliver alerts. Requires valid email address and SMTP credentials configured server-side. Returns success/failure status. Note: This operation is irreversible, verify recipient before calling.'
• Document output schema explicitly: { status: 'success'|'error', message: string, timestamp?: ISO8601, error_code?: 'invalid_email'|'auth_failed'|'timeout'|'network_error' }
• Add error handling guidance to description: 'On failure, returns error_code indicating root cause. Retryable errors: timeout, network_error. User-fixable: invalid_email (check format). Fatal: auth_failed (check credentials).'
• Add parameter validation rules: 'to' param should state 'Valid email address (RFC 5322). Example: [email protected]. If invalid, returns error_code: invalid_email with suggestion.'
• Implement confirmation pattern: Add optional 'dry_run' boolean parameter (default false). When true, validates inputs and returns preview without sending. Prevents accidental sends.