task_games__battleship__ship_status_count

Contract

1. Domain: games
2. Scene id: battleship
3. Public task id: task_games__battleship__ship_status_count
4. Supported query_id values: sunk_ship_count, partial_ship_count
5. Answer schema: integer_count
6. Annotation schema: bbox_set_map
7. Program schema: count(filter(ships, ship_status=target_status)); scene=battleship; scope=ship_status_count; query_branch=partial_ship_count

Program Contract

Program: count(filter(ships, ship_status=target_status)); scene=battleship; scope=ship_status_count; query_branch=partial_ship_count

Candidate set: the visible game board, pieces, tokens, cards, tiles, marked state, legal-move cues, result panels, and labeled options inside the ship_status_count objective scope. Operands: visible scene state and prompt-bound operands named by filter, ships, ship_status, target_status, battleship, ship_status_count, query_branch, partial_ship_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_map schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: sunk_ship_count, partial_ship_count.

Reasoning Operations

Families: filtering, counting

Generation Notes

1. Query ids are internal replay/sampling keys and do not define public task units.
2. Annotation maps each counted ship name to the hit-cell boxes on that ship, projected from the same generated game state used for answer verification.