| Domain | Owning article (revision) | M3 obligation | Outcome or remaining defect |

| Combat and contested work | Combat Rules `af22ead1ab5b…`; Combat and Enemy Roster `bb15ab8dce45…` | Combat-state contract and roster | **Checked.** Four records in the roster — **R3**, **R6**, **R8**, **R10** below. Combat Rules itself is clean. |

| Information, controls and accessibility | Player Information and Options `9ca194c21bb7…`; Player Information `bf126be1ebe0…` | Cue/action/refusal inventory and options | **Checked.** One clarification — **R9** below; the article already carries a current-room row, so no new state is proposed. |

**The selected opening, through both legal rig choices.** Objective 1 runs R01 → R02 → R03; the fasteners are at R03 and the loose Fork at R02. Both choices are legal and neither is an upgrade. *Deployed:* the board's own source supplies power and the Fork passes one board pulse to both compatible outputs, holding the rack and the patient-bay shutter together; the part leaves carried property and does not return on death. *Packed:* the selector must hold rack power throughout, so that enemy lane stays open, and the Fork comes home. **Corrected in this revision.** An earlier version said the freed bridge fits retrieval-ready but not air-heavy without dropping something. That inverted the actual constraint, and it was stated as a two-sided check, which makes it worse rather than better. Survival L54 and Mechanical35 L192 agree: retrieval-ready is 6 of 10, and 6 + Fork 1 + bridge 2 = 9 with one unit spare; air-heavy is 8 of 10, and 8 + bridge 2 = **10 exactly, so the bridge fits** while the Fork stays deployed. What cannot also be carried in that branch is the **Fork** — 8 + 1 + 2 = 11 — unless a canister or another object is deliberately left behind, and nothing is silently discarded. The Fork-in-rig branch also has its own control cost under Mechanical35: Manual Trigger moves to inactive cargo, so sensor-timed barrier behaviour replaces manual barrier timing.

**Shifted rediscovery.** R01 stays directly outside the airlock. The reveal moves to R01's outer unbound face, or to the nearest free frontier of the fixed ship-rooted group if that face is already latched. Checked in First Expedition Options and Player Information, which agree. The former omission in Player Information and Options was **R9**, now repaired in its existing row.

**Short rescue and home.** Prepared requires the route walked this life, the abort room stocked, an armed survivor-operable retreat and a canister at the route's authored full-attempt return mark. Sent commits only on that same person aboard with the ship airlock closed before the lethal boundary. Death mid-transfer: the armed conditional safeguard gives an alive/unsent world outcome at the named counted refuge, while the astronaut's knowledge stays unconfirmed because the closure tell is local. Checked in the continuity candidate, Rescue Rules, Cast31 and First Survivor, which agree. The older rescue shape and unconditional abort wording were **R1/R2**, now aligned; the unadopted comparison status was **R12**, now dated explicitly.

**Late station connection and ending dependency.** Objective 12 is physical completion under the specified provisional joining-family/bridge proposal; the final joining choice and completed-station behavior are not owner-selected. It does not depend on objective 10, rescue, treatment or any ending, and completing it neither triggers nor blocks a departure. Checked in Regional25, Connecting the Station and Story30, which agree, and all three keep stabilisation conditional.

The author register reconciles this assessment, the Systems 552 audit and critique559. Its rows group multiple passages and overlapping articles; the former eighteen-passage/severity total was not a consistent unit and is not used for acceptance. The final PM inventory below records eleven design articles plus two planning notes, all repaired and read back. The before/after wording below is dated review evidence, not a request for more work.

| **R1** *(PM-found; the most consequential record)* | Rescue and Home `a7f1e300f1e6…` L44 **and** Rescue Rules `a8e15548bf54…` L26–30 | Both describe the rescue as a "finite sequence of prepared refuges", with an intermediate shipward vestibule and a separate final one | Cast31 `61ff0b968eba…` L7 allows "at most one exposed junction, one airlock-side abort vestibule and the refuge door"; L45 has both D1 routes "converge on the same present abort vestibule, **not a refuge chain**"; L75 names D1 "the short correction of Rescue And Home". The owner's constraint is that rescue not become a long escort | A reader can author or expect a multi-segment chain that the accepted short-crossing contract excludes — contrary to the accepted one-junction proposal used to meet the owner’s no-long-escort constraint | Replace the chain sentence in each with the single-vestibule form: *"One current short route — airlock corridor, at most one exposed junction, one stocked airlock-side abort vestibule, the refuge door. Interruption returns the person to that same vestibule; there is no chain of intermediate refuges."* Retain living-abort versus conditional-death, full-attempt air, the armed safeguard and retreat before geometry. Owners: Rescue and Home; Rescue Rules |

| **R5** *(audit 4)* | Story and Ending Map `03c9cb8ee955…` L144, and L133/L140/L143 | "12 placed mechanisms"; "the future cast map must decide"; "up to 5 of the existing 15 scenes/variants to allocate" | Encounter33's thirteen with P13 the cross-listed core-isolation instance; Cast31's completed five-variant allocation; Room32's placements | Twelve to thirteen naming P13; mark the cast-map rows as allocated by Cast31, keeping exact occurrences later. Owner: Story and Ending Map |

| **R6b** *(audit 3)* | Two Language Discovery Map `df0f48f5bd39…` L123 | "The future 40-room/12-mechanism ledger must assign…" | Room32 and Encounter33 supply first-pass placements at thirteen | State the ledger as supplied at thirteen; exact occurrences and strings stay later. Owner: Two Language Discovery Map. *(Language29's other "twelve" uses are objective counts and are correct — not records.)* |

**Disposition of the grouped findings.** PM repaired eleven design articles and qualified two dated planning notes. This includes the linked rescue geography/control/retry clauses and the distinction between resolved conditional world outcome and unconfirmed player knowledge in Information34. The same work required no owner answer, new game content or additional ordinary review cycle. Exact final revisions are recorded below.

**Owner preferences short of a requirement** (recorded so they are not silently promoted or dropped): equipment loans at home, which the owner kept with an explicit hedge rather than a decision, and practical clue-based exploration, which the owner accepted as a direction with observation and survivor context helping interpretation. The first version of this register omitted both; neither changes a disposition, and neither is a settled rule.

| 5. Protected retreat versus injury/loss stakes or bind-first? | **No** — the protected-retreat candidate is specified conditionally; the alternatives are outlined choices, not three accepted detailed contracts | **No, with conditional wording** — the baseline states the branch as conditional rather than asserting one death-in-transfer outcome | Yes: rescue authoring and encounter geometry |

No answer is inferred from silence or from availability messages. Question 5 is the only one that touches M2 at all, and it constrains the *wording* of the baseline — conditional rather than asserted — rather than blocking one. An earlier version marked it "Partly", which read as a block that the same paragraph then denied.

**M2 — coherent conditional core baseline: ACHIEVED.** The opening, reusable builds, carrying/oxygen, death/shift, short rescue and home contracts now agree at the inspected first-pass interfaces. The PM integrated the evidenced discrepancies after the single critique/response; open owner choices remain labelled branches rather than assumed answers.

**M3 — complete first-pass design wiki: ACHIEVED.** All fifteen domains have their required first-pass behavior/inventory and interface evidence. The thirteen parent patches were guarded against their exact preceding revisions and read back byte-for-byte. The two domains marked Linked in the author matrix were separately read and verified by PM at scope37 closure, including the six packet dependencies; the present planning-note repairs were also read in full. Completion is based on those sources and repairs, not on agreement or the number of pages.

**What remains after the repairs**, and it is not small: the M4 design work — full scripts, every string and glyph occurrence, exhaustive state and layout permutations and admissible-arrangement specifications for all forty rooms — plus implementation and gameplay evaluation in a separately authorized later phase. Those are visible, owned and out of M3 scope by prior decision, not gaps discovered here.

**Dated author effort and calendar recommendation, before PM completion.** My first estimate of 1–3 role-hours was too low: it counted four records in four articles, and transfers plus guarded readback across eleven stopped articles is the real cost. The reviewer's 3–6 is closer; with R1 added across two more passages I would say **4–7 role-hours** for the repairs, and this response is already spent. That remains inside the 10–18 forward allowance. On scope alone M3 is therefore **not work-constrained**, and a **provisional 10 September** readiness date is my reasoned estimate rather than a forecast change — F1.14 stays PM-owned until disposition. M2 needs nothing further from this assessment, and the two decisions are independent: M2 is core coherence, M3 is whole first-pass readiness, and neither establishes fun, balance, usability, performance or duration.

**Corrected coverage statement.** The matrix now records **thirteen Checked and two Linked** domain rows. An earlier version said nine and six in its prose while the matrix showed eleven and four, and it listed traps as link-only although its row was Checked; "older superseded layers" was also counted as a domain, which it is not — it is a class of article, not one of the fifteen. Equipment/acquisition and colony/home were the two genuinely unchecked obligations named in the critique; both were checked two-sided for this revision and their results are in the matrix. The two that remain **Linked** are production scope and development handoff, both verified by resolved reference and internal consistency only. I did not re-review any closed domain's design merit, re-derive counts from first principles, or examine every permitted arrangement. The Systems 552 audit was unavailable for the initial draft and was incorporated in the ONE response. The final PM integration also used the independent factual verification of that response and all thirteen patch payloads. Nothing here establishes that any of this is discoverable, usable, balanced or enjoyable; those need later owner-authorised implementation and observation, and none has occurred.