77 lines
7.3 KiB
JSON
77 lines
7.3 KiB
JSON
{
|
|
"schema": "original-delivery-unit-v1",
|
|
"plan_id": "P23-03",
|
|
"title": "生活の観察",
|
|
"authority": {
|
|
"approved_plan": "docs/plans/original-completion-2026-10-02/IMPLEMENTATION_PLAN.md#p23",
|
|
"original": "docs/traceability/original/BASELINE.txt",
|
|
"original_sha256": "75e4bdbdeb379c9269553f0c776bdb9cc96e01f2d6aa5ca650d132c63a8aeb1a",
|
|
"original_lines": [
|
|
10,
|
|
13,
|
|
16,
|
|
25,
|
|
37,
|
|
43,
|
|
69,
|
|
83,
|
|
106,
|
|
216
|
|
]
|
|
},
|
|
"original_work": "部屋、行列、EV、職員、物資、営業状態が断面上で理解できる表示へ調整する。",
|
|
"original_acceptance": "人物や設備の表示が実状態に対応し、同需要の動線改善を観察できる。",
|
|
"denominator": {
|
|
"packages": 30,
|
|
"delivery_units": 88,
|
|
"this_unit": 1,
|
|
"partial_credit": false
|
|
},
|
|
"commander_direction": {
|
|
"path": "docs/plans/original-completion-2026-10-02/execution/p23-03/OPUS_CONCRETE_IMPLEMENTATION_DIRECTION.json",
|
|
"run": "83e836e232c74975aaf51f4fab5a1af9",
|
|
"codex_model": "gpt-6-astra"
|
|
},
|
|
"conditions": [
|
|
"room_occupancy_and_operation",
|
|
"queues_and_overflow",
|
|
"elevator_state_and_passengers",
|
|
"staff_state_and_jobs",
|
|
"supplies_and_carried_goods",
|
|
"same_demand_route_improvement",
|
|
"normal_os_and_distinct_pid_resume"
|
|
],
|
|
"scope": [
|
|
"All original six categories remain required: room, queue, EV, staff, goods, business state. Generic display of actual definitions/state, never an implementation limited to fixture IDs.",
|
|
"Representative acceptance scenarios use existing original-matched constructible definitions and real runtime operations, no undefined facility proxies or invented costs/mana/staff capacities. They do not establish all30 construction, finished asset, rights or natural-gameplay acceptance.",
|
|
"Retain P08/P22 package dependencies in plan. Available unit inputs: integrated P08-02/03, P22-01 inventory, current Runtime/report definitions and existing Kenney/code art. Unavailable P08-01 geometry/production values, D08/D09 quantities, P22-02/03 asset work remain explicitly open; they may not be silently considered done.",
|
|
"No Runtime/service/production data values or new assets. No visual redesign or new UI entry. Necessary existing renderer/read-only presentation wiring and evidence only."
|
|
],
|
|
"worker_deliverables": [
|
|
"Phase0 API_AND_STATE_MAP.md: real file:line and SHA for same-seed/reset/arrival inputs/build commands/clock/public reports/formal save-load APIs. Confirm actual deterministic demand mechanism before related implementation. If unavailable, preserve work and report precise missing capability and minimal separate proposal; do not manufacture same-demand proof or stop unrelated preparation.",
|
|
"One coherent frame observation model consumed by actual drawing, with public deep-copy get_life_observation_report. Cover actual room users and occupancy vs queue, state reasons, queues/overflow, EV position/load/direction/phase, staff task/state/layer/position, carried goods and actual supply report data. Preserve existing motion/layer/nonmutation/selection behavior.",
|
|
"Visible per-room business reason using FacilityInstance enum names; distinguish closed/no access/no power/no staff/degraded and suspended raid/damage from actual fields. No ambiguous reliance solely on tiny color lamp or utilization badge.",
|
|
"Visible normal-view route density and actual supplies, reusing established visual style and existing get_cell_traffic/report fields; no fictional stock/request data.",
|
|
"tests/ui/test_p23_life_observation.gd, diagnostics fixtures, Root run_p23_life_observation.py and strict checker with meaningful negatives, condition matrix and honest implementation metadata. No placeholder PASS or fixed all-true assertions."
|
|
],
|
|
"evidence_requirements": [
|
|
"Whole typed actual state and draw-consumed report rows; checker independently recomputes each relevant field using enum/schema definitions, never trusts all-boolean or self-asserted report. Screenshots pixel/rect provenance must correspond to actual rendered report positions.",
|
|
"LayoutsA/B: identical recorded seed, demand/arrival script, budget, initial clock, service definitions and unchanged production settings; actual Runtime paths differ only through predeclared allowed layout improvement. Record actual arrival input events, paths/arrival times, queues/wait metrics/density/completed services. Predeclare relevant metric; B strictly improves without new arbitrary numeric threshold. Fixture-only fabricated state is not valid same-demand proof.",
|
|
"Root actual OS observation on both layouts across six categories, FRONT/BACK, ordinary pan/zoom; formal save and distinct actual PID normal load, state/display remains correct. Handler tests or submitted scripts are not OS.",
|
|
"Collector request/ACK lifecycle, full actual CUA tool call IDs/arguments/screenshots/calibration from start, named helper self-properties, exact PID/engine/source/spec/settings/environment and guarded whole default-player profiles. No partial failed run splicing, no raw receipt relabel.",
|
|
"Use existing owned process/ordinary audio teardown gates. Check engine end/cleanup/protection on every run. Preserve failure history.",
|
|
"Sparse diagnostic snapshots with every acceptance field, content-hash reuse and streamed manifests; no unnecessary all-state-per-frame copies or arbitrary review byte limit. Required source/evidence may not be omitted for size.",
|
|
"Final exact frozen candidate: worker static/local and impact first; Root native/OS then mandatory one full regression, independent Opus all-condition review, integration/final/ledger. Reuse only matching actual source/input/settings/environment applicability.",
|
|
"Pin stable immutable spec/contract/fixture/engine/source plus relevant normative plan acceptance text. Record progress metadata separately, avoiding unnecessary reruns from unrelated progress-only plan text changes. Never ignore normative input change.",
|
|
"No Windows same-distribution/device credit, P23-02 held-gesture or x3 performance credit, P25 finished asset/full motion credit, or all30 facility credit from this unit."
|
|
],
|
|
"worker_instructions": [
|
|
"Read original work/acceptance, actual recorded Opus direction and relevant current sources before changes; do Phase0 static API audit first.",
|
|
"One Codex Astra worker, no nested providers/agents/parallel workers. User removed all individual repair/inference/material caps; do not stop on old2/3/500k restrictions. Timeout900 unchanged. No API fallback/new billing/permissions.",
|
|
"Root only executes Godot/native/OS/full regression after genuine submitted. Worker runs meaningful local Python syntax/checker tests only, no Godot or full regression. Missing Root evidence is expected at worker submission; state it explicitly.",
|
|
"Keep exact allowed paths. completion_marker_contracts append only one strict source authority entry; suite-policy only if new suite uses existing ordinary cleanup, sourceSHA-specific entry preserve all old policies.",
|
|
"Use existing run_all_tests.py/source refs/strict marker/protected-profile/typed codec/owned process infrastructure read-only; no broad harness reimplementation or proof-gate removal.",
|
|
"If actual required defect outside allowed scope occurs, provide failing evidence/minimal proposal and preserve independent work. Root and Opus decide scope next."
|
|
]
|
|
}
|