You would want to use a wildcard for that, combined with a 'Run an Application' action.
For instance, you would want to have a command with a spoken phrase of 'search for*'. The idea is that whatever you say after, 'search for' will be the search term. You can access that value by using the token {CMD_AFTER} (which is the portion of the spoken phrase AFTER the asterisk (*)). If you execute a 'Run an application' action using that token, you should be able to get close to what you're after (see attached images).
I say, 'close', as the Windows speech engine is not exceptionally adept at picking up words/phrases that are not predefined. That is, if you define your spoken phrases ahead of time, the speech engine is very good at recognizing what is said. If you are using wildcards, the speech engine (which is not very contextually tuned) is going to make best guesses as to what you are saying. That might end up being a little frustrating at times (your mileage may vary).
Hope that helps!