Towards Automated Fact-Checking of Real-World Claims: Exploring Task Formulation and Assessment with LLMs