task_games__chess__target_square_attacker_count

Contract

1. Domain: games
2. Scene id: chess
3. Public task id: task_games__chess__target_square_attacker_count
4. Supported query_id values: king_square_attacker_count, white_piece_attacks_target_square_count, black_piece_attacks_target_square_count
5. Answer schema: integer_count
6. Annotation schema: bbox_set
7. Program schema: count(attackers(marked_target_square, queried_side)); scene=chess; scope=target_square_attacker_count

Program Contract

Program: count(attackers(marked_target_square, queried_side)); scene=chess; scope=target_square_attacker_count

Candidate set: the visible game board, pieces, tokens, cards, tiles, marked state, legal-move cues, result panels, and labeled options inside the target_square_attacker_count objective scope. Operands: visible scene state and prompt-bound operands named by attackers, marked_target_square, queried_side, chess, target_square_attacker_count plus the active query_id branch. Operation: evaluate count 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 integer schema; generation binds a unique final answer. Annotation witnesses: annotation uses the bbox_set schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: king_square_attacker_count, white_piece_attacks_target_square_count, black_piece_attacks_target_square_count.

Reasoning Operations

Families: counting, spatial_relations

Generation Notes

1. Annotation marks bounding boxes for all attacking pieces from the queried side.
2. For king_square_attacker_count, the marked target square contains the king and the queried side is the opponent.
3. For the white/black target-square queries, the marked target square is empty.