task_symbolic__agent_automaton__future_grid_label

Summary

1. Domain: symbolic
2. Scene: agent_automaton
3. Task id: task_symbolic__agent_automaton__future_grid_label
4. Goal: simulate a turning-agent automaton and choose the option showing the full grid after the requested number of updates.

Program Contract

Program: agent_automaton.future_grid(scene=agent_automaton, scope=source_grid_and_future_grid_options, rule_variant=binary_rule|three_state_rule, output=option_letter)

Candidate set: the four visible future-grid option cards. Operands: the source grid cell states, start arrow cell and direction, selected rule variant, and requested step count. Operation: simulate the turning-agent rule for the requested number of updates and select the unique option whose full grid state matches the result. Output binding: answer is the selected option letter. Annotation witnesses: a bbox_map with source_grid and selected_option roles. Query ids: single.

Reasoning Operations

Families: state_update

Contract

1. Query ids: single
2. Rule variant metadata: binary_rule|three_state_rule
3. Answer type: option_letter
4. Annotation schema: bbox_map
5. Annotation target: role-keyed bboxes for source_grid and selected_option
6. Scene variants: clean_grid|lab_panel|notebook_grid
7. Generation controls: board size 3..5 rows/cols, step count 3..6, grid options 4
8. Prompt bundle: symbolic_agent_automaton_v1
9. Source file: src/trace_tasks/tasks/symbolic/agent_automaton/future_grid_label.py

Annotation Contract

The annotation is a bbox_map with two roles:

1. source_grid: bbox around the visible starting grid with the initial agent marker. The source panel is marked with START; the marker is not part of the source-grid annotation bbox.
2. selected_option: bbox around the correct future-grid option card.

The answer and annotation are bound from the same simulated execution trace.