task_symbolic__music_staff__key_signature_label

1) Identity

1. Domain: symbolic
2. Scene id: music_staff
3. Task id: task_symbolic__music_staff__key_signature_label
4. Objective contract: key_signature_label

Program Contract

Program: music_staff.key_signature_label(scene=music_staff, scope=key_signature_in_staff_context_plus_visible_text_options, output=option_letter)

Candidate set: the visible text option cards for key names. Operands: the visible key-signature accidental pattern and the key name represented by each option. Operation: identify the key signature from the positioned accidentals and select the unique matching option. Output binding: answer is the selected option letter. Annotation witnesses: the scalar bbox of the visible key signature. Query ids: single.

Reasoning Operations

Families: matching

2) Scene + task contract

1. Entities/relations: a rendered music-staff notation panel with marked notes, chords, measure ranges, option cards, or key signatures depending on the task objective.
2. Supported query_id values: single
3. answer_gt.type: string
4. Default annotation_gt.type: bbox
5. Annotation schema: bbox
6. Annotation witness policy: annotation marks the visible key-signature bbox; visible text options, context notes, staff lines, and decorative background are not prompt-facing annotation.
7. Rendering contract: the key signature is rendered as positioned treble-staff accidentals in standard order, not as a plain # # or b b text string.
8. The correct semantic key name is rendered as one visible option; answer_gt.value is the option letter.

Query id	User-facing operation
single	The prompt asks the one stable objective contract for this task.

3) Prompt contract

1. prompt_bundle_id: symbolic_music_staff_v1
2. scene_key: music_staff
3. task_key: music_key_signature_label
4. Query keys: internal branch key from the generated trace; public single-operation tasks expose query_id=single after registry normalization.
5. Output modes: answer_only, answer_and_annotation

4) Determinism + constraints

1. Deterministic generation and rendering from instance_seed.
2. Answer and annotation are bound from the same finalized notation scene and projected bboxes.
3. Count tasks construct the requested answer count before rendering; label tasks construct a unique target label by scene state.
4. No semantic constraints are auto-relaxed after sampling.

5) Source files

1. Task source: src/trace_tasks/tasks/symbolic/music_staff/key_signature_label.py
2. Scene shared package: src/trace_tasks/tasks/symbolic/music_staff/shared/
3. Config: src/trace_tasks/resources/configs/domains/symbolic/music_staff.yaml
4. Prompt asset: src/trace_tasks/resources/prompts/symbolic/music_staff/symbolic_music_staff_v1.json
5. Focused tests: tests/test_symbolic_notation_tasks.py