torii-simulatelisted

# Torii Simulate Use this skill as a compact knowledge base for operating and reasoning about SUMO experiments. The agent may decide which tools, workflows, references, or direct inspections are useful for the task. ## Common Knowledge - Separate simulator setup, controller behavior, generated outputs, and scientific interpretation. A run can succeed technically while still giving weak evidence for the intended claim. - For controller comparisons, paired route, demand, seed, horizon, signal constraints, and metric definitions make differences easier to interpret. - Completion, insertion backlog, teleports, warnings, and action traces often explain misleading performance metrics. - When debugging, preserve as much of the scenario as practical and change the smallest useful thing first. This is a heuristic, not a mandatory execution order. - TraCI is an interface to the simulator. What a controller can read through TraCI is not automatically equivalent to a deployable physical sensor system. - Controller objectives, pressure, reward, prediction error, and traffic outcomes are different quantities and should not be conflated. - Use the simplest evidence that answers the user's question. A full workflow is unnecessary for a narrow question. ## Reference Library Read only the references that help with the current task. | Topic | Reference | |---|---| | Environment and executable checks | `references/preflight-sumo-environment.md` | | Torii CLI/MCP routing | `references/mcp-t