Compare layouts.
Plan a better depot.
The original depot exposes a traffic problem. A planning-tool brief asks it to help a reader make a decision, which requires a comparison they can trust.
Edited build brief
Keep the depot engine and collision rules. Build a local layout-comparison tool for someone choosing between two depot designs. Let the reader save a named layout and workload, clone it into a trial, and run baseline and trial from the same recorded start. Offer a fixed number of ticks per comparison. Report delivered parcels and waiting ticks, show unfinished work, and export the two configurations with their results. Clearly separate editing, running and reviewing states. Verify repeatability and reject invalid imported layouts.
How it feels
The reader tries a design, compares a result and makes a choice. The moving couriers provide evidence for that decision.
What the agent needs to build: interface
Scenario names, clone, baseline/trial views, a run length and a result comparison replace an endless animation as the central journey.
What the agent needs to build: structure
The agent needs separate scenario and runtime records, deterministic resets, accumulated wait counters and a comparison report. Reuse the routing engine and give each run its own mutable world.
Acceptance check
Compare a layout with its unedited clone for 50 ticks. Results should agree. Add a wall only to the trial; confirm the baseline record and starting couriers are unchanged.
Design exercise: try a new build from this edited brief. Status: proposed; implementation and testing pending.