pipelines/CRITIC_fortnite-liveops-funnel.md
fortnite-liveops-funnel (fresh-context refutation)Draft under review: LENS_fortnite-liveops-funnel_DRAFT.md (15 findings, MISSING ×8 / WEAKER ×7).
Critic method: every citation opened; every zero re-run by me with my own positive control;
MISSING vs DESIGNED-UNBUILT re-derived independently; dedupe checked against
docs/DESIGN_GAP_REGISTER.md v1.0/v1.1/v1.2 (read in full, 463 lines) and
docs/PRE_5090_BUILD_PLAN.md; divergence guard honored (the F2P season economy is
pre-registered DIVERGENT-BY-DESIGN — any finding re-litigating it dies on sight; none did).
Standing note: this is an unusually disciplined draft. It self-refuted 9 candidate findings before
writing (§"Refuted and dropped"), and every one of those refutations I spot-checked is correct.
My kill list is therefore short and my amendments are mostly citation-hygiene, not substance.
---
Citations verified line by line:
docs/FUN_REBUILD_PLAN.md R4.3 — quote is VERBATIM ACCURATE (opened; the R4.3 bullet carries"today B03 asks the player to 'attune the breath' (a vril action) before B04 teaches vril gathers
back", the Ember/Stone Skin/Tremor ordering, and the Cloaking-is-the-evade-tool fun-critic note).
docs/factory_contract.json — I re-ran the term counts myself against the raw file: teach 0 · tutorial 0 · onboard 0 · first time 0 · introduc 2 · positive control
festival 5, difficulty 1, FR-0 86. The author's zero is REAL and my control matched, so
the search could have hit. (Nit: 86 FR-0 occurrences ≠ 67 riders — the "67 riders" figure comes
from docs/FACTORY_CONTRACT.md, not from an FR-id count; harmless, the zero is what carries.)
docs/DESIGN_GAP_REGISTER.md entry 32 verified: "Ability rails hollow: 153/185 abilities lack anunlock chapter … (Rank 13.)" — the data-emptiness claim is the register's own row, correctly cited
and correctly NOT re-reported as a new gap.
Verdict class check: WEAKER is right, not MISSING — the discipline demonstrably exists (R4.3 is the
proof) and what is absent is its generalization. Reuse-first: yes (one rider on an existing register
docs/proposals/QA_WATCHING_PROGRAM.md §1.5 funnel list verified at L110-114: RecordAttackPhase, RecordPlayerAnswer, RecordInputLatency, RecordHitStop, RecordStagger, RecordCameraSample
— every one a combat site, as claimed. QS-0 row verified at L349: "The ONE per-frame record schema
(JSONL, one vocabulary, 4 KB) … lands before any C++". QS-13..18 verified at L362. T2-6 verified at
L371 ("Charter schema + finish-the-chapter + break-it"). Per-frame clock vocabulary verified at
L44-45 (frame_index, t_game, dt_app, dt_game, time_dilation, t_wall) — so the proposed
progression group genuinely sits beside an existing group.
docs/PRE_5090_BUILD_PLAN_VOL2.md QT-17 (L1420) and QT-19 (L1422) verified verbatim in substance:QT-17 is "the measured budget vector read from the DTs", QT-19 gates "content ≥ the chapter-band
floor" — content inventory, not traversal. Density model as completion-hours verified (L34/L1418).
RecordCameraSample|RecordHitStop" Source/` returns zero in the game repo; positive control
PlayHitStop returns 3 hits in HumanityCombatFeelSubsystem.cpp. So the whole telemetry lane is
spec-only. That makes the *funnel family* DESIGNED-UNBUILT — but the finding is not about the
funnels' existence, it is that no progression field is in the SPEC at all, which I confirmed by
reading §1.5 and the §5 table. MISSING is the correct class for the instrument class. SURVIVES.
to gate a tuned number. That matches VOL2 L1422's own firewall language. Good citizenship, no
amendment.
is SessionId" is garbled — SessionId is not a time field (verified: HumanityQACaptureSubsystem.h
L292 FString SessionId, and L5 documents the capture path). The *point* (the telemetry lane is
design, not code) is independently true per my grep above. AMEND the sentence, keep the finding.
Every load-bearing fact independently verified in C:\dev\Humanity\Humanity:
Source/Humanity/Public/Discovery/HumanityClueSubsystem.h L1-12 — header quote is VERBATIM,including "NOT a journal (HL_0070/HL_0096: the journal does not open until Ch 13)" and "the held-key
HUD overlay lists the last 5".
TMap<FName,FString> CluePlayerTextById, TSet<FName> FiredClues ("a clue's sensation appends once"), TArray<FString> Sensations (ring), SeededChapter idempotence
guard. No Save/Load member exists — I read the whole header.
Source/Humanity/Public/Quest/QuestSaveGame.h — persists EXACTLY the five fields claimed (ChapterId, CompletedBeats, LocalFlags, ChosenForks, LastChosenForkId) plus
SaveVersion = 1. The proposed bump to 2 lands on a field the header itself says is "carried for
future migration".
Source/Humanity/Private/Core/HumanityCharacter.cpp Input_QuestLog / Input_QuestLogEnd (held-key field-sense overlay, "NOT the journal (which opens at Ch 13)").
The draft's line numbers are ~L1262-1283 in the file I opened — accurate.
grep -rln "SaveGame" Source/ → QuestSubsystem, WorldStateSubsystem,their two SaveGame headers, QuestSubsystem.h, WorldStateSubsystem.h, and the quest automation tests.
The seam exists and is used twice; the Discovery lane is absent from it. Zero is real.
BUILD-NOW — L876 and L1302 both state "QuestSaveGame has no scene-seen ledger". CN-6 covers
scenes. Nothing in CN-6, the WU wave clause (L892: "a save→quit→Continue cycle restores
mid-chapter"), or the register covers the clue/sensation lane. So this is not a re-report of
CN-6 — and the draft says so itself. SURVIVES, and I would raise it to the top of the HIGH four:
it is slice-scoped, built-code-adjacent, hours of work, and it defects the one continuity surface
the pre-Ch-13 player owns.
*"MISSING ASPECT: player expression verbs (emote/pose/photo/showcase) — distinct from
presentation/identity"* to COMPARATOR_LENS_PROGRAM.md. That string is not in that file (I
grepped it; positive control fortnite returns L57/L115/L141 in the same file, so the search
matched). The real source is a prior cluster return in the wave scratchpad (ret_08.json L98) and
its landed form is docs/DESIGN_GAP_REGISTER.md §v1.2 new-aspect rows:
"player-expression-verbs (roblox-fortnite)". The nearest in-file support is
COMPARATOR_LENS_PROGRAM.md L72-73 cell 24: "Photography / observation mode … only a P3 lens
touches expression verbs." AMEND the citation to the register + L72-73. Substance unaffected.
"ac-traversal (sightseeing/vista aspect) + roblox-fortnite (P3 expression verbs)", state
CLAIMED-RUNNING in part (P1-B). ac-traversal-stealth-legibility is running in THIS wave — its
draft sits beside this one. LOF-12 must be reconciled against that draft's photography finding at
fold time or the register gets two rows for one cell.
docs/DESIGN_GAP_REGISTER.md entry 22 is real and is exactly what thedraft claims ("Four T1 docs are byte-level EMPTY stubs, not drafts: Audio_Spec, UI_UX_Spec,
Personal_Dimension_Spec, Tier_B_DLC_Catalog. (Ranks 20/24 + brief.)").
recommendation is a site-appropriate verb set driven by care_tier data, explicitly "never a lock",
and it names the timid read and rejects it. That is the calibration CLAUDE.md asks for. No
care-inflation here.
---
docs/spine/CH_PROLOGUE.md "Mechanical anchors" → abilities_unlocked bullet: quote isVERBATIM, including "the four SOVEREIGN faculties are given in full as the ceiling tutorial
(Sovereign Vril Projection, Rift Sealing, Temporal Sight, Familiar Communion — T3_Core §6.4) and
are STRIPPED at the Ch-1 rift-fall … no persistent unlock carries past the fall (the ceiling is
the reference, not a kept kit)".
T0_Ability_Tree_Registry [ACTIVE v0.1] opened: 185 rows, 35 columns, **no lineage/descendantcolumn** — confirmed by reading the header list, not by grep.
A_E3_002 verified in-row: ability_name = Vril Sight, tier_name = Zelator, `tier_numerical = 2.0, unlock_chapter_ref = Ch_15`. The draft's "Ch_15 Zelator T2" is exact.
(Prologue|sovereign) over the whole canon repo → 0 files. Control: Rift Sealing` over the
same repo, same tool → 11 files (CH_PROLOGUE, T3_Core_Characters, Boss_Encounter registry,
Quest_Definition registry, Chapter_Index, transitions, STORY_REEL, two slice records). The search
could match and did. Zero is real.
PRE_5090_BUILD_PLAN.md L88: "Ability registry data pass (unlockchapters ×153, costs, cooldowns, classes)" — no lineage field planned, exactly as claimed. And I
re-counted the ×153 myself against the real column name (unlock_chapter_ref, not
unlock_chapter): 153 blank / 32 filled of 185. Exact.
what is under-specified is the lineage declaration. Reuse-first: one column on a registry a ranked
item is about to populate. SURVIVES, HIGH, unamended.
docs/spine/DECISIONS_PENDING_JOSH.md axis 6 (L245-249): quote VERBATIM — "THE EVENT LAYER —Hollow-driven LARGER EVENTS at every scale: SOLO … TEAM (once the MMO releases), RANDOM world
events, and SET-SCHEDULED seasonal or weekly challenges."
HOLLOW_EXPANSION_INTEGRATION_2026-07-27.md §1.6 verified at L279-290: the GREENFIELDpositive-control paragraph ("Nothing in T1, CVD or T0 defines a rotation") and the tooth ("Every
scheduled event row declares a changed_law_ref, and a window reusing the immediately preceding
window's law FAILS the gate") are both verbatim.
grep -rl "changed_law_ref" registries/ → zero files. The nearest candidate registry, T0_Hollow_Codex [DRAFT v0.1], holds 1 row and
14 columns (`hollow_id, manifestation_designation, spawn_locus_type, region_id, chapter_window,
contain_verb, purify_harvest_fork, sigma_trigger_anchor, perception_gate, reveal_tier,
containment_reward, firewall_anchor, care_register, source_ref`) — **no window/schedule/repeat
column**. So "names a column on a row in a registry that does not exist" is exactly true, and I
checked the one registry that could have falsified it.
registries/T0_Festival_Registry [DRAFT v0.1]/Sheet1.csv: 16 rows; columns exactly as listed;
window_kind = calendar_annual ×6, event_triggered ×5, pawukon_cycle ×3, lunar ×1,
daily ×1; player_participation = opt-in-participate ×11, observe ×4, assist ×1. **Every
number in the draft is right.**
recommendation executes the Hollow record's OWN ruling, verified at L1289 — "No Hollow event
borrows, overlays, or imitates a real living festival". Executing a landed care ruling, not
inventing timidity. Correct as written.
window_spec carries "timed to the agricultural year (SHAPE only, exact window lands with the LW-3 runtime)"; FST_0008/0011/0012 pawukon_cycle;
FST_0009 lunar (FST_0009 = "Nyepi eve + Ogoh-ogoh and the day of stillness"). Read from the CSV.
T1_Tier_C_MMO_Spec [ACTIVE v1.0] **§7.3 is literally titled "WorldState Modulation"** (L243) — verified against the heading list (§7.1 Regional Vril Modulation /
§7.2 NPC Biography Variation / §7.3 World State Modulation / §7.4 Tier B Catalog Protection /
§7.5 Hard Line Chapter Protection).
reach 100%". SURVIVES — but no independent register row (see structural amendment).
non-repeatable|once-only|once_only|unrepeatable over docs/` → 2 hits, both ordinary narrative
prose (docs/review/STORY_REEL.md, docs/spine/CH_58.md), neither a design class. Control:
weekly challenge|seasonal event|live-ops|SET-SCHEDULED over the same corpus, same tool → **5
files** (PRE_5090_BUILD_PLAN, DECISIONS_PENDING_JOSH, the Hollow record, a quest-seed queue page, a
book extraction). The recurring half matches; the once-only half does not. Zero confirmed.
FOMO search matches*Fomorians*. Good method hygiene.
cleansings accumulate on a separate seasonal tally that never enters the completionist
denominator". The FOMO objection really is pre-answered by a landed law, so the draft's
self-refutation-the-other-way (#9 in its dropped list) is the correct anti-timidity call.
masteries and progression and easter eggs" ARE gated on cleansing-meter progression. The draft's
guard is right, but it holds only if the *threshold* stays offline-reachable while only the
*spectacle* is once-only. Say it explicitly in the column definition or a later reader reads the
two rulings as contradictory. SURVIVES with that sentence.
T1_Tier_C_MMO_Spec §8.9 (L345-351, heading "**§8.9 Single-Player Party Building as Raid TrainingGround"): the quoted sentence is VERBATIM** — "Tier C carries the highest meta-collapse
pressure of any arena because server communities converge on shared solutions, so the two canonical
counterweights are the per-server strategy variation at §8.5 … and the sweeps, which measure
whether the plurality is real."
item pool rotation|pool rotation|rotate the pool|vaulting over docs/ → 2 hits,both the word "vaulting" in a Uluwatu *sacred-site traversal* note. No item/ability-pool rotation
exists. Real zero, independently derived.
timidity — I verified the doctrine in the same §8.9 paragraph: "mathematically overpowered builds
are allowed, meta-collapse is the banned failure mode, and the target is many roles and strategies
with hundreds to thousands of equally viable builds." Correctly refused, correctly reasoned.
aggregate sweep can hide") is sharp and costs two sentences. Keep it. SURVIVES.
in Tier C World Bosses, territory control, and competitive PvP."
entry point 0 · onboard 0 · starting state 0 · character creation 0 · new player 1 · control guild 94. The doc
is searchable and the zero is real. (Draft said guild 66; true count 94 — the control fires
either way, but correct the number.)
PRE_5090_BUILD_PLAN.md L151-157, including "with the selection surfacebecoming the returning-player skip path" and "The mechanical half is cheap NOW and expensive after
players have saves." The timing argument is real, not rhetorical.
scope ruling** — the base game's netcode posture (single-player-first? always-online? how the MMO
layer attaches). GAS OnRep hooks imply a latent decision." The draft does not cite it. A P2.11
entry-architecture line that assumes a Tier-C-only account pre-empts item 27. Write the line as
a *shape reservation* ("the entry surface must not foreclose a third, non-campaign entry state"),
not as a commitment to always-online. With that wording it stays a zero-ruling item.
adaptive-AI weighting magnitudes, and sweep thresholds" to Tier C build phase, and a cold-start
spec could be read as riding that. I sustain MISSING because the cold start is **not in the
enumerated deferral list** and §3.3 states the promise present-tense. Flagged CONTESTED.
T0_Minigame_Registry [DRAFT v0.1] = 2 rows. MG_0001 Caci (whip-and-shield honour duel),
tier marquee, engine_family duel-parry, region flores_island, chapter CH_03, care_tier
elevated; MG_0002 Senet, structured, board-of-fate, CH_15. 19 columns — **no
festival_link, no arena_scope, and boss_link/situation_link are empty on both rows**.
T0_Festival_Registry FST_0004 Caci contest season, window_kind = event_triggered,
gameplay_hooks = "the whip-and-shield contest circuit; seats the player's first off-hand Shield
(WPN_033) and grounds the CH_03 main-boss PARRY verb; climb for community standing and reward,
offered never forced; gong-waning scores the rounds" — verbatim. Same activity, same region,
same chapter, two registries, no key. The central observation is factually exact.
docs/proposals/systems/minigame-galaxy.md §5 schema verified at L133-150: culture/region_id, chapter_id, boss_link, situation_link, link_rows, plus "A mis-named FK fails the harness
loudly".
that file "returns nothing." Not quite — arena does match once, at L117 ("the arena/gladiator
marquee (Ch 28) is framed as coercion the player fights *out of*"), an unrelated sense. State the
zero as *no festival/event/Tier-C routing field or reference*.
docs/ROADMAP_POST_5090_TO_SHIP.md L422-425 verified verbatim: "Steamworks integration … : achievements, cloud saves over the existing WorldState SaveGame stack (PRE_5090_BUILD_PLAN.md
P2.7), the overlay, and controller/input via the built Enhanced Input config (P2.3)." P4 exit
verified at L452. EA boundary verified at L15/L441.
T1_Combat_System_Spec L300 and L336 verified verbatim ("cosmetic achievements", "withintegrity-band variants").
(L63) already run an in-world achievement economy: of the fifty capacity pieces, "the remaining
thirty [are] carried by agility-course puzzle variants, achievements, and world secrets." Two
consequences: (a) the "derive, don't author" recommendation gains a fourth existing ladder; (b)
there is a naming collision to rule at P4 — an in-world "achievement" (a capacity piece) is not
a Steam achievement, and the derived-list emitter must not conflate them.
integrity_score integer is gone from thesurface") is real; the denominator rule + offline test are real. SURVIVES.
docs/MARKETING_WEB_PROGRAM.md verified: "## The four lanes" at L24 — 1 Capture, 2 Asset,3 Web skeleton, 4 i18n. All four are launch-asset lanes, exactly as claimed.
cadence 0 · patch 0 · update 0 · community 0 · roadmap 3 · discord 1. The draft's "returns only the positive-control
note … and one pointer to P4's store page" understates it slightly. State the counts. The substance
— no post-launch update-rhythm lane, no published promise, no owner — holds.
and it costs nothing. Keep it.
and is honestly priced LOW. Consider routing it to MARKETING_WEB_PROGRAM.md as a fifth-lane note
rather than spending a register row. SURVIVES, lowest value in the set.
§16.2 "Mode variety per chapter" (L893), §16.3 "Anti-staleness mechanics" (L901). All three
exist and are exactly what the draft says — arc-scale and chapter-scale, never session-scale.
PRE_5090_BUILD_PLAN_VOL2.md L34, L1418): "≈100-200hmain / 1-2k h side+weave / 8-10k h 100%".
DESIGN_GAP_REGISTER.md L41,"| 9 | Rest / safe-site session-anchor loop | wow | WEAKER | P2 / W-SPACE | No* |", and L297 records
the Elden Ring Sites-of-Grace merge into it. The draft declines to re-report the anchor half and
reports only the *read* half — exactly the dedupe discipline the wave rule asks for. Reported-never-
gated, so the tuning firewall is respected. SURVIVES, low priority, honestly priced.
art time, which is the right trade for a solo build."
layer is an authoring obligation that RECURS FOREVER once live-ops starts … a seasonal calendar
authored against a marketing cadence generates player-facing vocabulary indefinitely, and each new
season needs its own §17.1 read."
*law* lane, which has the same shape — is correct, and it is the sharpest single observation in the
audit. The fix ("no repeat within N windows over a finite versioned enum" rather than "never reuse
the previous law") is reuse-first and one sentence. SURVIVES.
---
1. LOF-1 (reuse target, MATERIAL). The recommendation names gate 30 factory_contract plus a
rank-13-dependent join. There is a cheaper, already-ARMED host the draft missed:
harness/check_spine_graph.py L328-340 enforces **produce-before-consume over local.* keys,
ordered by declared beat seq** — E(f"{bid}: consumes {k} before any prior beat produces it") at
L331, plus a WARN on gate expressions at L340. It is registered in harness/gates_config.json
L297 and, per harness/gates_scorecard.json L78, it already runs over all 79 spine nodes
(CH_PROLOGUE … CH_EPILOGUE). So the teaching-order rider needs neither rank 13 nor a new join:
model the teaching beat as the producer of local.taught_<verb> and the demanding beat as its
consumer, and the existing armed gate catches the exact R4.3 inversion class today, on every
chapter, at authoring time. Re-point the recommendation there; keep the gate-30 rider as the
declarative half.
2. LOF-1 (figures + a stronger zero). "67 riders" is not derivable from the json (FR-0 occurs
86×) — source it to docs/FACTORY_CONTRACT.md. And EXTEND the zero: I ran it on the prose contract
too — FACTORY_CONTRACT.md gives teach 0 · tutorial 0 · onboard 0 · teaching order 0 ·
unlock order 0 against control rider 46. Both halves of the contract are silent; say so.
3. LOF-2 (garbled control). Delete "the only per-frame time field in QACapture is SessionId"
(SessionId is an FString path token, HumanityQACaptureSubsystem.h L292/L5). Replace with the
control I actually ran: `grep -rn "RecordAttackPhase|RecordInputLatency|RecordCameraSample|
RecordHitStop" Source/ → 0, control PlayHitStop → 3 in HumanityCombatFeelSubsystem.cpp`.
Same conclusion, sound evidence.
4. LOF-9 (pre-emption guard + count). Cite open-decisions item 27 (Tier-C netcode posture,
PRE_5090_BUILD_PLAN.md L1965) and word the P2.11 line as a shape reservation that does not
commit to always-online. Correct guild 66 → 94.
5. LOF-10 (zero restatement). arena matches once in minigame-galaxy.md L117 (the Ch-28
gladiator marquee, unrelated sense). Restate as: no festival/event/Tier-C routing field.
6. LOF-11 (missed citation + new constraint). Add T1_Combat_System_Spec L63 — thirty of the
fifty capacity pieces already ride "agility-course puzzle variants, achievements, and world
secrets" — and add the naming-collision clause (in-world achievement piece ≠ Steam achievement).
7. LOF-12 (MANDATORY citation fix). The "MISSING ASPECT: player expression verbs
(emote/pose/photo/showcase) — distinct from presentation/identity" quote is NOT in
COMPARATOR_LENS_PROGRAM.md. It is a wave-scratchpad cluster return (ret_08.json L98) whose
landed form is docs/DESIGN_GAP_REGISTER.md §v1.2 new-aspect row "player-expression-verbs
(roblox-fortnite)". In-file support: COMPARATOR_LENS_PROGRAM.md L72-73 (cell 24: "Photography /
observation mode … only a P3 lens touches expression verbs"). Re-cite. **This is the only citation
in 15 findings that fails the open-the-file test — everything else held.**
8. LOF-13 (honest counts). roadmap ×3 and discord ×1 do appear; cadence/patch/update/
community are 0. State the real numbers.
Structural amendment (register hygiene, at fold time): LOF-6, LOF-7 and LOF-15 are all clauses on
a registry that does not yet exist (LOF-5's T0_Event_Window). Landing them as four register rows
gives the register four rows for one unminted schema and guarantees three get re-discovered as "no
owner" later. Fold them as LOF-5 with three named clauses — (a) real-clock vs in-world clock
disjoint window_kind sets, (b) repeat_class {recurring|seasonal|once_only} with the spectacle-only
guard, (c) changed_law_ref bounded to a finite versioned enum with a no-repeat-within-N rule. One
entry, three columns, one owner.
---
Looked specifically for care-inflation, defanging, and omission-as-safety across all 15:
it"), rejects it on the rated-M bar, and rules that "the care work is data … not prohibition; the
answer is a site-appropriate verb set, never a lock." That is the CLAUDE.md calibration exactly.
landed line (L1289), and it *enables* the event layer rather than restricting it.
one-time events *because* FOMO is bad would have been the timid call, and the denominator rule +
offline test already answer it.
"mathematically overpowered builds are allowed, meta-collapse is the banned failure mode".
serve. Nothing to flag.
The F2P season economy is pre-registered DIVERGENT-BY-DESIGN (COMPARATOR_LENS_PROGRAM.md L57,
verified verbatim: "fortnite-liveops teaches F2P season economics, divergent by construction").
No finding re-litigates it. The draft isolates only separable craft halves — the legible finite
ladder without the purchase (LOF-11), update rhythm without the storefront (LOF-13), encounter-law
rotation without player-pool vaulting (LOF-8), scheduled spectacle without FOMO pricing (LOF-7) — and
pre-registers all four exclusions in its own DIVERGENT section. Nothing to kill here. Other standing
guards also checked: no-scoreboard (LOF-11 respects CVD §17.12's minigame top-score exception
explicitly), markerless/diegetic navigation (untouched), register rows 9/22/32 (cited, not
re-reported).
docs/DESIGN_GAP_REGISTER.md (v1.0 top-10 + v1.1 rows 20-45 + v1.2 board)QuestSaveGame has no scene-seen ledger", L876/L1302) — LOF-3 isthe clue lane, not the scene lane, and the draft says so. Not a duplicate. This was the single
most likely kill in the set and it survives on the evidence.
roblox-fortnite (P3 expression verbs) | CLAIMED-RUNNING in part (P1-B)") — **the one live
collision**: ac-traversal-stealth-legibility is running in THIS wave and its draft sits in the
same directory. LOF-12 must be reconciled against it at fold time or cell 24 gets two rows.
Director action, not a kill.
---
Killed: 0 · Survived: 15 · Amended: 8 · Contested: 4
**Landing recommendation: LAND, after the 8 amendments — amendment 7 (the LOF-12 mis-attributed
quote) is MANDATORY, and amendment 1 (re-point LOF-1 at the already-armed produce-before-consume
gate) is the highest-value change in this pass.**
This draft survives refutation better than any I have reviewed in this wave. It self-killed 9
candidate findings before writing, every remaining citation but one opened correctly, every zero I
re-ran independently reproduced, and its two strongest findings (LOF-3, LOF-5) are verified down to
the field list in built C++ and in live CSV. Landing order:
continuity surface the pre-Ch-13 player owns. I would rank it above LOF-1.
check_spine_graph.py's armed enforcer — it then costs a namingconvention rather than a rank-13 dependency, and it protects Ch 3-13 while they are being authored.
register; LOF-14 as a derived-column note on QT-17.
Contested, for the director:
1. LOF-12 × cell 24 — live overlap with the ac-traversal draft in this same wave; dedupe at fold.
2. LOF-9 class — MISSING vs DEFERRED-BY-PHASE. I sustain MISSING (the cold start is not in the
spec's enumerated deferral list and §3.3 states the promise present-tense), but it is the closest
call in the set, and its P2.11 line must not pre-empt open-decision item 27 (netcode posture).
3. LOF-7 × axis 7 — axis 7 gates abilities/masteries on the cleansing meter; the once_only guard
holds only if the threshold stays offline-reachable and only the spectacle is once-only. One
explicit sentence in the column definition closes it.
4. LOF-13 scope — a marketing/business finding inside a design register; honestly priced LOW and
date-free, but it may belong in MARKETING_WEB_PROGRAM.md directly rather than as a register row.