What problem does it solve? Running Scenic scenarios against CARLA fails in confusing ways because sampling and simulation are separate phases, and errors like empty blueprint categories, unsatisfiable map filters, or client/server version mismatches all surface as opaque runtime failures. This Skill separates the phases, diagnoses each failure class, and guarantees bounded runs that restore world state afterwards. ## Core Features & Use Cases - Headless pre-flight sampling: Compile and sample .scenic files without a simulator to distinguish syntax problems from map problems before spending time on a full simulation. - Bounded, verified execution: Run scenarios with --simulate --2d plus count/time bounds, then confirm a simulation actually terminated rather than trusting the exit code. - Environment diagnostics: Check that Scenic and the CARLA client live in one interpreter, that blueprint tables match the client version, and that the server is reachable. - Use Case: A user reports that carlaChallenge10 rejects every sample. The Skill inventories each map's road features, shows the scenario needs an unsignalized four-way intersection the current map lacks, and retargets it to a suitable map with --param overrides. ## Quick Start Ask the agent to run the Scenic scenario carlaChallenge1 against the running CARLA server as a smoke test of the setup.
