✠ The Chronicle

✠ ENTRY XLVIII

`adjust-npcs` debuts and closes the content-lifecycle set

all five /adjust-* stewards have now logged a pass — while two owner sessions fix a red-hand starter deck and rebuild the /dev surface end to end. The nightly breadth check stays red for a second consecutive night on the same already-diagnosed regex bug (still unshipped, not new). A fresh reduced-nightly measurement reads a lower late/higher impossible terminal rate than 2026-09-05 — directional only.

CONTENT

adjust-npcs pass 1 — the fifth and final content-lifecycle category debuts

What changed

Staged the Forest Ranger and the Lost Trader, unblocking the get-to-cave route.

Why it changed

The 2026-09-05 digest flagged npcs as the one /adjust-* category still reporting "never." This pass closes it — cards, equipment, enemies, keywords, and npcs have now all logged at least one pass since the per-item content-lifecycle split (Phase 132-era, 2026-09-03).

commits f0a2891f · a6d6679d

MECHANICS

Owner session: tri-colour starter deck, stop de-duping the card base

What changed

STARTING_CARD_IDS grew from two BODY-only cards to one card per stance (BODY/MIND/HEART), and buildCombatDeck stopped collapsing every authored copy count to one.

Why it changed

A live combat playthrough (instrumented combat-round-e2e) found two independent defects: the ensureStarterCards fallback dealt a 100%-BODY hand off a two-card deck (HEART/MIND dice dead on arrival), and de-duping meant the 18-card Threadbare Office was dealt as an 8-card deck — below the deck-removal floor. Re-measured the stage-profile matrix at matched settings before/after: byte-identical, because that harness builds decks from unique ids and never saw the de-dup repeal.

commit 3fb4963b

INFRA

Owner session: /dev surface audited and rebuilt

What changed

Retired nine stale/duplicated debug controls (a retired loot-rarity model, dead encounter buttons, raw-setState playthrough presets, seed/Aporia/map-reset buttons, synthetic dialogue/quest fixtures) and added a live STATE inspector, world travel to any map+node, COMPLETE MAP, labyrinth acts I-III, an any-foe enemy picker, reward triggers, real NPC dialogue/quest pickers, and buff/ debuff/item pickers. CI-facing testIDs were preserved.

Why it changed

Phase 132's extraction kept a stale, duplicated /dev route pending a later audit; this pass makes every late-game surface (reward channels included) reachable in a tap.

commit e3807ae8

CONTENT

adjust-cards pass 2 — 2 created, CHAIN/OMEN second carriers

What changed

Authored "The Village Comes Over the Hill" (rank 4/Rib, heart) and "The Ducking Stool" (rank 3/Splinter, body) as second carriers for CHAIN and OMEN, each of which had exactly one carrier per /adjust-keywords pass 1's rider-inclusive audit.

Why it changed

Both keywords anchor a load-bearing atlas family (the damage octet; tempo-and-control) rather than being accidental cruft, so the loop-call called for a second carrier, not a retirement.

commits 8c346ac7 · 4113dc0d

CONTENT

adjust-equipment pass 2 — zero-action re-audit

What changed

Re-ran the full structural sweep over relics/consumables/ shop pools; nothing needed shipping this pass beyond a ledger sha fix.

Why it changed

Standing rotation cadence — a clean re-audit is itself a signal (no drift since pass 1's five dead-consumable rewires).

commits 721adac9 · 5ff070d7

BALANCE

Baseline re-measured — directional stage read

What changed

The SessionStart hook flagged the standing baseline (42cc0412, reduced-nightly) stale by 3 mechanics-source commits. Re-measured via the reduced-nightly pass (30 runs/cell) at commit 5ff070d7. Stage terminal win rates across the 144-cell matrix: early 97.3%, mid 95.4%, late 73.9%, impossible 35.4%. Overall combat-quality index 0.73 (per-stage 0.69/0.70/0.76/0.74 — read-only, not a governing score).

Why it changed

This digest's own freshness gate exists so a balance question never gets answered against a stale engine, and this window carried two real mechanics-adjacent touches (the starter-deck/de-dup fix, two new CHAIN/OMEN cards).

Note: Late and impossible read notably different from the 2026-09-05 measurement (86%/26% → 73.9%/35.4%) — a 12pp late drop and a 9pp impossible rise. This is a single-seed, reduced-confidence read; not cited as a regression or filed as a /deck-tuning candidate from here — measure and report only, per this skill's own rail — but it's a large enough swing that a confirming full 3-seed baseline:regen looks worthwhile before anyone treats it as settled.

commits 5ff070d7 · baee · ee · this · c

✠ The marginalia

The night's watch — what ran while nobody looked
when (UTC)verboutcome
2026-09-05 15:19–15:53march → adjust-npcssuccess — pass 1, Forest Ranger + Lost Trader staged
2026-09-05 18:22–18:25marchsuccess, no-op
2026-09-05 ~20:48 (PR #288)owner sessiontri-colour starter deck + stop de-duping the card base
2026-09-05 23:57–00:02marchsuccess, no-op
2026-09-06 01:07–01:16march → critiquesuccess — pass 32, 0 findings
2026-09-06 02:36–02:42marchsuccess, no-op
2026-09-06 04:31–04:37marchsuccess, no-op
2026-09-06 ~06:13 (PR #289)owner session/dev surface audited and rebuilt
2026-09-06 06:40–06:44marchsuccess, no-op
2026-09-06 08:32–08:38marchsuccess, no-op
2026-09-06 10:26–11:14march → adjust-cardssuccess — pass 2, 2 created, CHAIN/OMEN second carriers
2026-09-06 12:35–12:38marchsuccess, no-op
2026-09-06 14:26–14:36march → adjust-equipmentsuccess — pass 2, zero-action re-audit
2026-09-06 16:28–20:32march ×3success, no-op
2026-09-06 22:25–22:34marchsuccess, no-op
2026-09-07 01:07–12:41march ×7success, no-op

All 20 march runs this window succeeded — no crashes, no cancelled runs. Both non-loop changes (the starter-deck fix, the /dev rebuild) shipped through owner-directed PR sessions, matching the pattern the last several digests have noted.

The measure — did the fighting change?

None from this digest's own meta-loop rail this window — the pulse shows a healthy content-lifecycle rotation (all five categories now active) and no starved queue or mistuned gate signal. /expand last ran 2026-09-03 (pass 11, 3 candidates); no new pass fell in this window.

What is still unfinished
  • Build plan (plan/steps/01_build_plan.md): 0 open phases ([ ]),

11 partial ([-]), 7 explicitly [skipped], 124 shipped through Phase 78 — unchanged this window.

  • plan/AUDIT.md: 67 rows under "Pending" excluding 46 resolved-but-

unmoved [x] bookkeeping rows (a housekeeping gap, not new this window) — includes the newly-updated breadth-check row above.

  • plan/CRITIQUE.md: 29 rows under "Pending"; last full pass

2026-09-06 (pass 32, 0 new findings, fully clean re-confirmation).

  • plan/PHASE_CANDIDATES.md: 52 rows under "Pending", unchanged —

last /expand pass (11) was 2026-09-03, before this window opened.

  • plan/CONTENT_LEDGER.md: cards + equipment now pass count 2;

enemies + keywords + npcs at pass count 1 — every category active.

  • No open triage:needs-user or loop:do issues.
  • Deploy: clean — HEAD is a docs/plan-only tick, no gated workflow

triggered.

  • Nightly breadth (e2e:minigames): RED, second consecutive night

— see Needs you.

  • Measurement: was STALE at session start (baseline 42cc0412, 3

mechanics-source commits behind); re-measured this tick via npm run baseline:regen -- --runs=30 --confidence=reduced-nightly → now FRESH at 5ff070d7. 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.