060 cloud Linux native functional QA / 2026-10-02 22:26–22:30 UTC
Agent-operated native input via CUA. Not a human enjoyment test. No audible audio verified.
App SHA-256: 19d922a0a9561bdfbcaa4e0b176d8005d7fa8303befa4d0e2b92c743e78047c7
Model SHA-256: 54a1f2a4dc14217d6a1bda608c3e3f953fbce7653f5a0921fd3acc873783e846

Launched via observed Application Finder, bash launch.sh. Actual title: 060 小さな進化島. Read Japanese title/help and trait picker on native screenshots.
Selected B, held Space Start for850ms: remained day0/E6 after entering gameplay. Selected C(fur) in slot1, T(claw) in slot2. Held Space900ms: advanced exactly one day, E6.
P paused. Held/released Space700ms while paused: no day advance. P resumed; fresh Space advanced to day2/E6. Selected first slot, removed fur: E4, changed=True. Native screen showed harvest4, base3, upkeep1, paid morning cost2, forecast4.
Tried adding F on the same morning: one_change rejected, E4 and day2 unchanged. F12 checkpoint/native_rejected_change.png.
Four fresh Space presses reached day6/E8, outcome won, egg result visible. F12 checkpoint/native_survival_eggs.png. S saved records/B-01-v1_table.json; confirmation text visible over terminal outcome.
X from success returned day5/E6. Held R650ms returned B/day0/E6. Selected pond; two Space presses with no traits produced E2 then E-2/failed. Failure and X recovery instruction visible. F12 checkpoint/native_failed.png.
X restored a running state. 3 switched to C/day0/E6. B comparison restart, P help: explicit comparison upkeep0 marker visible above normal-reference values. F12 final native_events.json/native_runtime.png.
ESC closed the game. cua.listWindows verified 060 window absent. GUI slot released to parent.

Preserved original F12 runtime event JSON, separate checkpoints and screenshots. No test rewrites these files. Native logs record meaningful operations and state hashes; physical held-key durations are observations from CUA calls, not invented key telemetry.
Known display issue identified during this pass: failed result still showed the next-day forecast, though model input was terminal-blocked. Any correction/revalidation is documented separately rather than changing this original observation.
