task_games__circular_chess__target_cell_reacher_count

Contract

1. Domain: games
2. Scene id: circular_chess
3. Public task id: task_games__circular_chess__target_cell_reacher_count
4. Supported query_id values: white_piece_reaches_target_count, black_piece_reaches_target_count
5. Answer schema: integer_count
6. Annotation schema: point_set
7. Program schema: count(filter(pieces(target_color), target_cell in legal_destinations(piece))); scene=circular_chess; scope=target_cell_reacher_count

Program Contract

Program: count(filter(pieces(target_color), target_cell in legal_destinations(piece))); scene=circular_chess; scope=target_cell_reacher_count

Candidate set: the visible game board, pieces, tokens, cards, tiles, marked state, legal-move cues, result panels, and labeled options inside the target_cell_reacher_count objective scope. Operands: visible scene state and prompt-bound operands named by filter, pieces, target_color, target_cell, legal_destinations, piece, circular_chess, target_cell_reacher_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 point_set schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: white_piece_reaches_target_count, black_piece_reaches_target_count.

Reasoning Operations

Families: filtering, counting, state_update

Generation Notes

1. The blue marker identifies the target cell.
2. Pawns, check, checkmate, castling, en passant, and promotion are intentionally out of scope.
3. Annotation marks source-piece centers as pixel-space points.