FoxBear corrected boot: visible controls and ordinary Quit
Run 20261118-foxbear-corrected-boot-visible-controls establishes visible
app-specific movement, jump and apple action, followed by ordinary F12 Quit.
One new registered dx_foxbear original was launched. No additional original,
runtime repair, shared helper edit or replay was needed. Earlier scopes remain
closed; this run does not repair their missing evidence.
The tested frozen recipe is the published 6308e7a3b source graph with module
eb2395e1c115980dd78dbb646a26f755a4119a2a1e086942e715240e07c34bc6.
The EXE and ART retain their previously documented 2ca53c3b… and
ad719996… hashes. Existing 084802a67/6308e7a3b/2e9875240
native oldfail/newpass, identity/refusal, queued-expiry, independent UP,
TERM/KILL lifecycle and final locked-build proofs were reused. Relevant live
helper bytes matched the published graph. Mandatory stdole/TLBS/Git-font
actual VFS boot, whole PNG renderer and normal Worker preflight passed before
dispatch, with the exact previously tested source/index hashes
f5a12a79…/f92b9747…. Main and a real Worker read actual registry media
aliases. An initial preflight-only invocation wrote append-only evidence;
the subsequent invocation refused that evidence overwrite before any process
dispatch. Those files were preserved under cheap-preflight/, then the same
recipe passed and launched the single original.
All reviewed images are complete 1024×712 native frames. initial.png shows
the forest, fox, bear, platforms, log and HUD. right.png shows substantial
forest scrolling and the fox farther right after NUMPAD4. before-jump.png
shows the fox standing on the ground; jump-fresh.png shows it airborne with
spread limbs after NUMPAD7. apple.png visibly shows a red apple held at the
fox's mouth/hand after NUMPAD3. left.png shows the fox shifted left and
facing left after NUMPAD6. NUMPAD5 and NUMPAD2 were also consumed with
independent UP, but their individual stop/crouch consequences are inconclusive
in the selected moving scenes (the bear overlaps the fox in crouch.png).
No stronger claim is made for those two actions.
The retained-frame native-context route, original capture times, 15-second
game freshness, 500ms confirmations, 3000ms operation window, independent UP,
native identity checks and Quit reserve were unchanged. One proposed jump
using an older frame was refused locally as NOT_SENT before DOWN. A newly
captured, personally reviewed frame and a new intent then admitted the jump;
this was not UNKNOWN or a replay. Six gameplay DOWNs were consumed and all six
independent releases acknowledged. No pause, clock/state/score changes,
bindings or macros were used.
The seventh DOWN was reviewed ordinary F12 (wParam=123, HWND65537,
thread1). Its native USER-return receipt was consumed, then native PID2391332
exited 0, signal null, with actual wait and both EOFs. No cancellation or
forced cleanup caused that exit. The prearmed final release correctly records
OWNER_EXITED, not a physical UP ACK. Watchdog exit0 and both EOFs survive;
containment reports leaked:false, terminated:null. Minimum observed free
home disk was 2057089024 bytes; no RAM gate was used.
Evidence is in scratch/runs/20261118-foxbear-corrected-boot-visible-controls/,
with exact Git, relevant binary worktree diff, owned patches, unique module,
sealed-read audit, finite control receipts and selected named PNGs. Git remains
the source archive; no source snapshot, media copy, frame series or capture
bundle was created. No Python, Wine, sub-agents, Codex workers, boxes or
deployment were used by this task. Box ID/state/expiry/stop receipt are
null/NOT_CREATED/null/null. The worker verifies cleanup, publishes
result.json last, releases ownership and stops.