test: expand core coverage, starting with Game/index.ts extraction targets

[nodots]

Problem

Core test coverage is low, and it is lowest precisely on the code the Game/index.ts decomposition (#132) is about to move. Extracting untested methods means the test suite cannot catch extraction regressions where the risk is highest.

Measured baseline (COVERAGE=1 npx jest, 2026-07-15):

Scope	Stmts	Branch	Funcs	Lines
core (all files)	56.5%	40.9%	56.8%	57.3%
Game/index.ts	53.7%	45.6%	61.2%	55.5%

Uncovered ranges in Game/index.ts map directly to extraction targets:

Method(s)	Approx. lines	Extraction target module (#132)
roll internals	666–831	turnFlow.ts
switchDice	1016–1117	turnFlow.ts
executeAndRecalculate	~1542	turnFlow.ts
restoreState	1962–1993	lifecycle.ts
canOfferDouble / double	2038–2095	cube.ts
acceptDouble / refuseDouble	2115–2208	cube.ts
resign	2284–2329	cube.ts
confirmTurnWithRobotAutomation	2356–2377	robot.ts
undoLastInActivePlay / canUndoActivePlay	2467–2489	undo.ts

Goal

Raise coverage on the methods being extracted to a level where the extraction is verifiably behavior-preserving, then lift core coverage more broadly. Approach: characterization tests first — capture current behavior as tests before the code moves, so a passing suite after extraction is proof of equivalence.

Scope (phased, ordered to unblock #132)

Phase A — safety net for the decomposition (do before the matching #132 phase)

• cube.ts targets: canOfferDouble, double, canAcceptDouble, acceptDouble, canRefuseDouble, refuseDouble, resign — cover accept/refuse/resign outcomes, cube value/ownership transitions, Crawford interaction. (Blocks refactor: split Game/index.ts god file into cohesive modules #132 Phase 3.)
• undo.ts targets: undoLastInActivePlay, canUndoActivePlay, and the snapshot push inside move — cover push→undo round-trip restores exact prior moving state, empty-stack error, wrong-state error. (Blocks refactor: split Game/index.ts god file into cohesive modules #132 Phase 2.)
• lifecycle.ts target: restoreState. (Blocks refactor: split Game/index.ts god file into cohesive modules #132 Phase 4.)
• turnFlow.ts targets: switchDice (incl. the must-use-larger-die guard), executeAndRecalculate, roll branches. (Blocks refactor: split Game/index.ts god file into cohesive modules #132 Phase 5.)
• robot.ts target: confirmTurnWithRobotAutomation, handleRobotMovedState.

Phase B — broader core lift (after / independent of #132)

• Raise core coverage to 75% (statements, branches, functions, lines) from the current ~57% line / ~41% branch.
• Add a coverageThreshold gate of 75% to jest.config.js (currently coverage is opt-in via COVERAGE=1, no threshold) so coverage cannot regress silently. The gate lands once coverage clears 75% — adding it earlier would fail the suite immediately at the current ~41% branch. Interim option: ratchet the threshold up as coverage rises rather than one jump at the end.
• Identify and cover other low-coverage core areas surfaced by the full report (e.g. debugSingleGame.ts ~54%, and the sub-50% files in the full run).

Delivery

• One PR per phase (both the refactor: split Game/index.ts god file into cohesive modules #132 extraction phases and these coverage phases), for reviewability. Each PR keeps the suite green and, for refactor: split Game/index.ts god file into cohesive modules #132 phases, keeps the public Game.<method> API unchanged.

Notes / constraints

• Coverage is gated behind COVERAGE=1 in jest.config.js (reporters: text, lcov). Full report: run COVERAGE=1 npx jest from packages/core.
• Current baseline suite is green (448 passed / 12 skipped, exit 0) — new tests must not depend on that changing.
• Characterization tests should assert on real produced game state, not re-implement game logic.
• Part of the Game/index.ts remediation epic; coordinate phase ordering with refactor: split Game/index.ts god file into cohesive modules #132.

[nodots]

Tracked under epic #133 (Game/index.ts remediation), alongside decomposition #132 and as any remediation #96. Phase A characterization tests are sequenced to land ahead of the matching #132 extraction phase.