This PR adds the required `task_categories` metadata (`other`) to the dataset card, since AgentTell is a benchmark for behavioural side-channel leakage in browser-use agents and does not map to a standard ML task category. It also adds a link to the GitHub code repository in the README so the code and benchmark materials are easy to find from the dataset card. No license was added because none is specified in the paper or repository.
