torii-reportlisted
Install: claude install-skill Tarard/Torii-SUMO
# Torii Report
Use this skill as a compact knowledge base for interpreting and communicating completed SUMO evidence. The agent may decide which references are useful and whether a reporting task needs only explanation, a comparison, a formal report, or additional analysis.
## Common Knowledge
- Traffic outcomes such as delay, travel time, queue, throughput, stops, reliability, emissions, and priority-user effects are different from controller reward, pressure, objective value, estimator loss, or prediction error.
- Check whether compared methods used the same network, demand realization, routes, seeds, horizon, signal constraints, evaluation population, and metric definitions before attributing differences to the algorithm.
- Completion matters. Arrived-only averages can be misleading when one method leaves more vehicles unfinished or waiting for insertion.
- Keep simulator truth, modeled observations, oracle information, connected-vehicle samples, and physically deployable sensing separate.
- When methods receive different information, distinguish information advantage from controller or algorithm advantage.
- State exclusions, invalid runs, unfinished demand, and uncertainty when they materially affect interpretation.
- For traffic-control reporting, explain what information reached the controller, how it was obtained, what action the controller could take, and what traffic outcome followed.
## Reference Library
Read only the references that help with the current task