Finite acceptance RS1-RS8 from the preserved draft (byte-identical DRAFT.md),
pinned to the collector final Q and the integrated full9 receipt on db9d613.
Docs only, outside the product source map; nothing registered or run.
Original credit unchanged (22/88, 4/30).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Representative input (602 min, 200 guests) cannot start Operations
(Day1 06:00 only, population cap 100), so Ops ON uses the real new-game
opening and 200-guest Ops ON stays unmet. Pin the autoload/main startup
order, define render counts from the world_preview model, separate input
shortfall from natural change, split checker/static review/owned-run proof,
and make the collector self-contained (tracked run_all_tests.py and
owned_outer.py only). Budgets and original denominator unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Short Git-tracked spec_ref with the exact P28-01 work/acceptance, the P28
cross-cutting six and BASELINE.txt line references. Collector task contract
(codex/gpt-6-astra, independent Opus review) limited to execution/p28-01/
collector/**, outside the product source map, so the formal 192-suite
evidence stays valid. Nothing registered or run; original credit unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Cite the real Runner receipts (submitted a5d2491, integrated/final b7f6168,
9/9 conditions, low findings preserved) in a new ACCEPTANCE record. Keep the
x1 budget miss, x3/stall/render/Windows, the unused staff service shaft and
the four low findings as separate unmet items. Pinned p06-01/p14-02
acceptance JSON unchanged; the 46-suite supplement lives in the new record.
Sync P28-01 summary and gaps to the r02 measurements; next unit P28-01.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- P14-03 on 83b170a: formal 192/192 (22,198 checks), Root pairs 5/5,
representative perf r02 (collection/functional comparison established;
x1 ~827 ms exceeds the 4 ms budget in both versions), independent Opus
whole-unit review PASS. Per the skill contract it stays uncounted (21/88)
until Runner's own validate/review/integration/final; the registration
proposal and evidence-binding verifier are under artifacts/.../p14-03-runner-gate-r01.
- Correction notes on P06-01/P06-03/P14-02/P06-02: their regressions missed 46
roster-only suites; the defects found there are fixed in 83b170a and the
formal 192-suite run covers them. Completions and approvals are preserved.
- IMPLEMENTATION_PLAN header and progress_snapshot next pointers now show 21
units, P14-03 at the Runner gate and P28-01 next; history kept.
- Evidence matrix: 50% slowdown attributed to test_runtime_path_traffic,
injected trips and budget/scope wording per the review.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The formal collector's roster includes 46 suites the Makefile never ran;
the first formal attempt stopped on test_required_kitchen and a sweep found:
- v15 MIMI purchase saves could not be written once Runtime saves carried
mana_reserve (P06-01): the format detector now validates that section and
excludes it before the unchanged v15 allowlist, like special_peeping_rooms.
- test_required_kitchen's legacy14 exact-load comparison now maps the two
approved migrations (fresh A06 reserve at the restored boundary, elevator
snapshot v1->v2) after checking each against its contract, in Runtime key
order; every other byte must still match.
- test_runtime_defense_autonomy allowed Mana to change only by the A06 flow.
- test_p08_normal_placement_ops expects 25 palette slots (two warps); its
suite-policy pin is updated with flags unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- tests/s2/test_new_pid_pairs.gd runs each writer/reader save-restart pair as
two real child processes and keeps their complete output in the private
profile, since the script-level roster runs suites without arguments.
- Commit the Godot UIDs of scripts added this session (and the pre-existing
peeping-room scripts that lacked tracked UIDs) plus the legacy fixture's
thumbnail import, so a private import generates nothing undeclared.
- Repin suite-policy source SHAs for test_main_dashboard (A06 HUD, palette 25)
and test_runtime_transport (sustained outage helper); flags unchanged.
- Register the pair suite's completion marker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- ElevatorService snapshot v2 carries the shaft-wide wait; v1 saves (written
with every elevator dynamic_cost 0) migrate by deriving it on load, and any
v2 mismatch, including all zeros, is rejected atomically. A genuine save
from 50ceca0 is kept as tests/fixtures/p14-03/legacy_rush_v1.
- A guest aboard a cabin no longer decides, leaves or despawns until the
elevator sets it down at a real landing; previously a rider told to leave
vanished from a moving cabin.
- test_route_wait_choice now proves the rush detour and route loss with a
worker who actually walks, every shared rider's arrival cell, service
elevator and stair congestion, legacy migration and a new-PID restart.
- Register completion markers for the seven suites added this session and
bring the P14 ledger in line with P14-02's completion.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Riders queued for any floor pair in a shaft now delay every pair: the
elevator service publishes the rest of the shaft queue as each connector's
dynamic cost, refreshed on trip changes and each cabin minute, and restore
rejects saves whose wait disagrees with their trips. A two-shaft fixture with
real guests shows rush diversion, shared cabins within capacity, recovery,
route loss and warp-exit congestion.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Agent Control Runner