Role-based access control decides what each identity may do through the API. A Role lists rules, verbs on resources in API groups, inside one namespace; a ClusterRole does the same cluster-wide or for cluster-scoped objects. Neither grants anything until a binding names subjects: users, groups or ServiceAccounts. A RoleBinding grants a Role, or a ClusterRole, in its own namespace only; a ClusterRoleBinding grants a ClusterRole everywhere.
RBAC only adds permissions, so start narrow: the verbs a task needs (get, list, watch for reading; create, update, patch, delete for changing) on the resources it touches. The built-in ClusterRoles view, edit and admin cover common cases when bound per namespace.
kubectl create role, create rolebinding and their cluster versions build these quickly. kubectl auth can-i VERB RESOURCE --as=USER tests the result without logging in as anyone.
A narrow role, a binding, and a check that says no.