The parsing engine and using [TOKEN]s comes in real handy when designing in HTML, but honestly using tokens for the content panes is kinda clunky.
It's much easier to code the content panes directly into the HTML.
There's only this rule to remember:  The skin file must have one, and only one, area for content named "contentpane".  In other words, you can have as many content panes on your page as you can fit, but one and only one must have the name "contentpane".

So, you can do like this in your HTML if using tables:
<tr>
    <td id="LeftPane" runat="server" class="thatClass"></td>
    <td id="ContentPane" runat="server" class="thisClass"></td>
    <td id="RightPane" runat="server" class="thatClass"></td>
</tr>
<tr>
    <td colspan="3" id="BottomPane" runat="server" class="widePane">
</tr>
or like this if using CSS-based skinning methods:
<div id="ContentPane" runat="server" class="thisClass"></div>
<div=id="SomeOtherPane" runat="server" class="thatClass"></div>

Hope this helps...