task_games__ultimate_tictactoe__line_completion_move_label

Contract

1. Domain: games
2. Scene: ultimate_tictactoe
3. Scene id: ultimate_tictactoe
4. Public task id: task_games__ultimate_tictactoe__line_completion_move_label
5. Supported query_id values: x_winning_move_label, o_winning_move_label, x_blocking_move_label, o_blocking_move_label
6. Answer schema: string_label
7. Annotation schema: bbox

Program Contract

Program: label(filter(highlighted_board_empty_cells, line_completion_move(player, tactic_kind))); scene=ultimate_tictactoe; scope=line_completion_move_label

Candidate set: the visible game board, pieces, tokens, cards, tiles, marked state, legal-move cues, result panels, and labeled options inside the line_completion_move_label objective scope. Operands: visible scene state and prompt-bound operands named by filter, highlighted_board_empty_cells, line_completion_move, player, tactic_kind, ultimate_tictactoe, line_completion_move_label plus the active query_id branch. Operation: evaluate label over the candidate set using the visible game state, rules, legal moves, comparisons, counts, simulations, or option-selection constraints encoded in the program expression; generation enforces a unique final answer. Output binding: answer uses the string schema; generation binds a unique final answer. Annotation witnesses: annotation uses the bbox schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: x_winning_move_label, o_winning_move_label, x_blocking_move_label, o_blocking_move_label.

Reasoning Operations

Families: state_update

Generation Notes

1. Query ids choose X/O and winning/blocking tactic semantics.
2. Annotation contains the selected option-cell bbox.
3. Query ids are internal replay keys and do not define public task units.
4. scalar_annotation_checked=true