If you've never made AI or bots then I suggest you first start with a basic reactionary system just for the sake of experience. Something like, "Bot must get from A to B" while avoiding obstacles as they come across them to the best of its ability.
A GOAP system is as its name suggests, goal oriented action planning. This is not goal oriented action performing. This means you will have to have at least two separate AI systems. One will be the GOAP system which basically says "these things need to be done in this order" while you will have your other AI system actually figure out how to do these things one at a time. The how is the hardest part and where you would benefit from just designing some simple reactionary AI.
Now it has been years since I last designed a GOAP system, and that was also in a different engine. I basically made dummy structures to contain each "step" where each structure contained references to the structure prerequisites for the step as well as a base cost and modified cost (calculated based on certain in-game factors) of performing the step. When I used GOAP I would form a list of these structures, effectively the path of steps to some goal with minimum cost, and pass it to my second AI which would figure out how to do the actual step in the list. When the step was completed it would pop it off the list and start on the second. The two systems were completely independent of each other.
The AI just had access to these GOAP structures but it had to determine what they meant and how to accomplish it. There was no actual 'do this' coded into the GOAP structures.
As for actually performing tasks you may benefit from looking into state machines. You may need to design multiple states to accomplish any one step.
I don't know if any of my rambling will help you, but here's hoping.
