Preliminaries are prefatory turns and sequences that are used to alert co-participants to the incipient social actions in talk-in-interaction. Using multimodal conversation analysis, this study examines the videorecordings of paired role play interactions in an L2 interactional competence (L2IC) assessment setting. We describe how the participants jointly manage to enter the task by deploying diverse preliminaries through pre-pre sequences, pre-question sequences and pre-telling sequences as sequential practices, and story-prefaces as turn design features. In so doing, they project the main actions in mutually recognisable ways and collaboratively accomplish the instructed actions in the complementary role cards. We argue that preliminaries used for entering the task talk mark the primary point for the participants to display their L2 ICs in the focal task-based assessment context. The findings bring conversation analytic insights into L2 IC and provide implications for assessment interactions.