✠ ENTRY L
A busy two-day window: The Capital (map 5 of the northern continent) shipped and got its own content-lifecycle backfill within the same window, four UI legibility fixes landed, the mid-combat-crash test axis finally closed out, and the standing three-digest-red breadth check got fixed and is confirmed green again. Baseline re-measured fresh after 3 mechanics-source commits.
What changed
/forge shipped a new 9-node map (the-capital), opened a travel door from town-across-river's boss node, and paid off the three-map advisor-selection narrative thread planted at ncy-5 → cr-9 → tar-4.
Why it changed
town-across-river's own code comment flagged itself terminal with "no door onward yet" — the sharpest thinness signal in this tick's Step-1 audit across maps/events/art.
What changed
adjust-enemies pass 6 found the-capital's roster at 83.3% overlap with northern-city (over the skill's >70% ceiling) and created two capital-native enemies — The Stamper and The Underclerk, both debt-office archetype — dropping the reading to 62.5%.
Why it changed
flagged as future backfill work by /forge's own commit body when the-capital shipped reusing northern-city's roster wholesale; this is that backfill landing one tick later, the same cadence caverns/connecting-river/town-across-river each followed.
What changed
adjust-npcs pass 5 staged a second NPC (The Ribbon-Picker) at cap-5, on the existing gathering pool — The Capital had shipped with only one staged NPC (The Herald) across 9 nodes.
Why it changed
skill Step 1's "map with fewer than 2 staged NPCs" signal; the new NPC reads the pre-existing sweetheart-was-nominated flag, no new persisted state.
What changed
adjust-keywords pass 5 found the card editor's keyword-face projection had no case for deal, recoil, recoil_x, immolate, or purge_self — 60% of the 128-card library (77 cards) fell through to a generic glyph in the editor's own preview.
Why it changed
all five already had a live KEYWORDS entry; this was a pure mapping omission, not a new keyword.
What changed
adjust-keywords pass 4, adjust-npcs pass 4, adjust-cards passes 5 and 6, and adjust-equipment passes 5 and 6 all came back zero-action this window — reachability, pricing, duplicate, scale-ladder, aspect-thirds, relic/consumable/shop/signature signals all reconfirmed clean.
Why it changed
standing rotation cadence; a clean re-audit is itself a signal the library isn't drifting.
commits 9e423a34 · 6b365d01 · 35c4c56c · 2b02ff60 · 01629acf · da51e629
What changed
the initial exploration-map camera only centred the open-node focus centroid at fixed 1x scale, so a wide branch could still leave the outermost node off-screen. computeFocusTransform now zooms out (never in) just enough to fit the whole current+available bounding box.
Why it changed
closed a standing CRITIQUE.md [3.5] row — off-screen open nodes no-op'd silently on tap, with no visual cue that panning was needed.
What changed
the opening-hand preview's -32px card overlap covered most of each card's name; widened to -22px and wired the same CardDetailOverlay/onInspect pattern the live in-round hand uses.
Why it changed
closed a standing CRITIQUE.md [MED] row (score 3.5) — any name still covered is now one tap away from the full card.
What changed
the narrative-choice label and description were the same source string (description just re-cased), so both /dialogue and /event rendered every reply's text twice.
Why it changed
closed a standing CRITIQUE.md [3.2] row from an external-critique session-playtest finding.
What changed
the LOG toggle and the tutorial coach both anchored off a static HUD-height estimate; a full stance-check telegraph plus an active alt-win meter grows the real HUD past it. Both now anchor off a measured HUD height reported on every layout pass.
Why it changed
closed a CRITIQUE.md [MED] row — the toggle's near-opaque pill was painting over the telegraph's tail line, the same bug class as the earlier signature-rune-column HIGH.
What changed
extended combat-round-e2e with a fourth crash-strict axis — the level-up-out-of-victory path, where a won fight's granted XP crosses the next-level threshold and cascades a level-up mid-aftermath. Wired as e2e:combat-round:levelup, added to both mobile and mechanics verify gates.
Why it changed
the standing [HIGH] "mid-combat crash 30 seeded runs could not reproduce" row explicitly named this as its one remaining untested axis; the row's own condition for closure was "either the crash reproduces or the remaining axis ships." It shipped clean — the crash itself remains unreproduced on web.
What changed
upgradeable-dice-e2e.mjs's die-label regex still matched pre-spec-33 wording ("stance die"/"BOON face"); ported the same color/usable extraction combat-round-e2e.mjs already used. This digest's own step-3 breadth check (e2e:minigames) ran clean — 10/10 sandbox+live runs, no crashes — confirming the fix held.
Why it changed
three consecutive dated digests (09-05, 09-07, 09-09) logged this exact red for the identical, already-diagnosed cause; nobody had picked it up until this window's /audit cycle did.
What changed
critique pass 35 found no new findings. expand pass 12 filed 3 candidates (empty accessory-kind gap, dead zoneHas engine hooks, early-encounter difficulty smoothing); pass 13 filed none — every live signal already tracked.
Why it changed
standing rotation cadence.
What changed
session start flagged the standing baseline (e57ab0db, reduced-nightly) stale by 3 mechanics-source commits (The Capital ship + two content backfills). Re-measured via the reduced-nightly pass (30 runs/cell, 144 cells) at commit b1624da0. Stage terminal win rates: early 97.3%, mid 97.0%, late 82.0%, impossible 8.7%. Terminal-outcome totals: 3,012 victory, 735 mercy/capitulate, 20 concede, 553 defeat, 0 retreat (unchanged aggregate shape from the last read — the three new commits were content/enemy-roster work, not deck or engine changes).
Why it changed
this digest's own freshness gate exists so a balance question never gets answered against a stale engine.
Note: doctrine since THE BIG NUMBERS REWRITE (2026-09-02) governs no win-rate curve or CQI target — reported as directional evidence only, per this skill's own rail. plan/CRITIQUE.md's standing [HIGH] late-stage global collapse — all 10 presets 0.00 late row reads stale against this measurement (late is 82.0%, not 0%), but resolving stale CRITIQUE rows is /critique's job, not this digest's — flagged below, not touched here.
✠ The marginalia
| when (UTC) | verb | outcome |
|---|---|---|
| 09-09 10:30–10:42 | march → expand | success — pass 12, 3 candidates filed |
| 09-09 12:40–13:06 | march → audit fix | success — LOG toggle/telegraph overlap |
| 09-09 14:30–14:57 | march → audit fix | success — die a11y copy drift, part b |
| 09-09 16:31–16:43 | march → critique | success — pass 35, no findings |
| 09-09 18:34–18:38 | march | success, no-op |
| 09-09 20:28–20:57 | march → adjust-keywords | success — pass 4, zero-action |
| 09-09 22:26–22:34 | march → adjust-npcs | success — pass 4, zero-action |
| 09-10 01:01–01:15 | march → adjust-cards | success — pass 5, zero-action |
| 09-10 02:36–02:45 | march → adjust-equipment | success — pass 5, zero-action |
| 09-10 04:33–04:42 | march | success, no-op |
| 09-10 06:44–06:56 | march → adjust-enemies | success — pass 5, zero-action |
| 09-10 08:36–09:21 | march → forge | success — The Capital ships (Phase W6) |
| 09-10 10:29–10:37 | march → audit note | success — CRITIQUE.md rune-column row stale-fixed |
| 09-10 12:40–12:45 | march | success, no-op |
| 09-10 14:29–15:05 | march → iterate | success — mid-combat-crash coverage completed |
| 09-10 16:29–16:38 | march → expand | success — pass 13, no candidates |
| 09-10 18:32–19:03 | march → adjust-keywords | success — pass 5, glyph backfill |
| 09-10 20:27–21:06 | march → adjust-npcs | success — pass 5, The Ribbon-Picker created |
| 09-10 22:25–22:52 | march → audit fix | success — exploration camera off-screen-node fix |
| 09-11 01:00–01:19 | march → audit fix | success — hazard route-select fan fix |
| 09-11 02:37–03:01 | march → adjust-cards | success — pass 6, zero-action |
| 09-11 04:33–04:45 | march → adjust-equipment | success — pass 6, zero-action |
| 09-11 06:43–07:16 | march → adjust-enemies | success — pass 6, 2 enemies created |
| 09-11 08:36–08:57 | march → audit fix | success — dialogue reply-card echo fix |
24 of 24 march runs succeeded — no CI-infra blips this window, unlike the transient ENOENT failure two digests ago.
None from this digest's own meta-loop rail this window — 24 of 24 march ticks succeeded, the content-lifecycle rotation stayed active across all five categories, the standing breadth-check red got fixed rather than lingering a fourth digest, and expand ran twice with no starved-queue signal. No gate/cadence/ceiling mistuning observed.
plan/steps/01_build_plan.md): 0 open phases ([ ]),0 blocked, 125 shipped.
plan/AUDIT.md: 69 rows pending (117 headers under "Pending" minus48 resolved-but-unmoved [x] bookkeeping rows); 5 needs-user-call.
plan/CRITIQUE.md: 21 rows pending; last full pass 2026-09-11(pass 35, 0 new findings, ran 09-09 16:41 — see above).
plan/PHASE_CANDIDATES.md: 44 rows pending (up from 41 — expandpass 12 filed 3 this window, pass 13 filed 0).
plan/CONTENT_LEDGER.md: cards, equipment at pass 6; enemies atpass 6 (2 created); npcs at pass 5 (1 created); keywords at pass 5.
triage:needs-user or loop:do issues.triggered yet (grace window).
e2e:minigames): GREEN — 10/10 sandbox+liveruns, no crashes. First clean read after three consecutive red digests; see the [infra] card above.
e57ab0db, 3mechanics-source commits behind); re-measured this tick via npm run baseline:regen -- --runs=30 --confidence=reduced-nightly → now FRESH at b1624da0. See the [balance] card above.
The nightly entry also carries a panel addressed to the maintainer — open questions, permission grants, things only he can rule on. It is held back from this page: it is correspondence, not a report. The unfinished work it tracks is published above, in plain words.