task_games__solitaire__cascade_card_at_depth_label

Contract

1. Domain: games
2. Scene id: solitaire
3. Public task id: task_games__solitaire__cascade_card_at_depth_label
4. Supported query_id values: single
5. Answer schema: option_letter
6. Annotation schema: point
7. Program schema: select_option(card_at_visible_depth(column, depth)); scene=solitaire; scope=cascade_card_at_depth_label
8. Scalar annotation checked: true

Program Contract

Program: select_option(card_at_visible_depth(column, depth)); scene=solitaire; scope=cascade_card_at_depth_label

Candidate set: the visible game board, pieces, tokens, cards, tiles, marked state, legal-move cues, result panels, and labeled options inside the cascade_card_at_depth_label objective scope. Operands: visible scene state and prompt-bound operands named by card_at_visible_depth, column, depth, solitaire, cascade_card_at_depth_label. Operation: evaluate select_option 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 option_letter schema; generation binds a unique final answer. Annotation witnesses: annotation uses the point schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: single.

Reasoning Operations

Families: ranking, spatial_relations

Generation Notes

1. The prompt names a 1-based tableau column and a visible depth counted from the top of that column.
2. The visual options show card faces, and exactly one option matches the target tableau card.
3. Annotation is one point on the visible part of the target card in the tableau, not the option card.
4. Prompt wording comes from src/trace_tasks/resources/prompts/games/solitaire/games_solitaire_v1.json.