What problem does it solve? Aiming the CARLA on-screen camera requires manual spectator transform math and actor lookups, and the API offers no way to attach the spectator to a moving vehicle. This Skill resolves natural-language targets like "the ego" or "the Toyota Prius" to a specific actor and positions or tracks the camera in preset views. ## Core Features & Use Cases - Preset camera views: Chase (third-person), first-person/driver, top-down bird's-eye, and front views with tunable distance, height, and pitch offsets. - Live follow mode: Re-sets the spectator transform on every world tick to emulate attachment, since the CARLA API cannot reparent the spectator. - Actor resolution: Selects targets by id, role_name (e.g. hero), or type_id wildcard filter, disambiguating multiple matches by nearest distance. - Use Case: While debugging a scenario, ask for a "1st-person view of the Prius" and the Skill lists actors, resolves the filter, and follows that vehicle from the driver's seat for 30 seconds. ## Quick Start Ask the agent to give you a third-person chase view of the ego vehicle and it will check prerequisites, resolve the hero actor, and follow it with the spectator camera.
