✠ ENTRY V
The loop drained four build-plan phases (9-12) end to end while the owner ran the largest parallel push yet: PR #41's 10-deck rebalance merged and grew into five more win-path-scaling items from Battle Lab round 2, then a three-PR, ~60-commit sweep retired the last of the pre-Hazard-Pattern "skill" combat system (save-migration, the Skill stat axis, the dead CombatState.log befriend cluster) down to a single remaining meaning: SignatureSkill. Actions minutes hit 90% of the monthly quota and got throttled mid-day. Critique's playwright-permission block is now a fifth consecutive recurrence. Breadth check: all five e2e suites pass.
What changed
verified the first real character/story/world specs (C-01-the-sophist, W-01-aporia-labyrinth-continent, S-01-fishing-village-northern-forest-dilemmas) already satisfy this row and recorded the phase as shipped.
Why it changed
each spec folder's template defines "the first real spec is *-01-*.md" — all three already existed pre-tracking; the phase closed without new authoring.
What changed
added Jest full-provider mounts + a Playwright web flow pinning the ExplorationScreen + (tabs)/_layout seam (state/e2e/cross-screen-integration.engine.test.tsx + scripts/exploration-combat-roundtrip-e2e.mjs).
Why it changed
closed the cross-screen regression blind spot flagged in plan/CURRENT-STATE.md — hermetic per-screen tests were blind to the FIGHT-modal-unmount class of bug.
What changed
a guided intro for the Rest encounter (The Night Watch), the smallest of the four unguided minigames.
Why it changed
GAMEPLAY_GAPS GAP-001 — no minigame had a guided first-time intro; hazard/cache/quest tutorials queued as follow-ups.
What changed
collapsed the orphaned pre-monorepo axiomancer-mechanics/spec.md duplicate to a pointer, fixed docs/combat.md's dead-API references (basePower/chipHp/DIRECT_DAMAGE_WEIGHT/GOLD_CARD_IDS) against spec 32 v3, and documented the previously-undocumented rank/rarity/ cardType axis.
Why it changed
the engine doc-drift AUDIT row is chronic — new surfaces outrun the docs every few weeks; this closed the current backlog.
What changed
the 21-agent card-rebalance PR merged after verify caught and fixed two real regressions pre-merge (debuff_kindling_ember missing from the v3 support-tag registry; an inverted alignment-authoring assertion) plus a Grace swayCapitulates() follow-up fix.
Why it changed
direct response to the URGENT "combat challenge-gradient retune" candidate — early/mid/late win rates were pinned near 1.00; this moved them to 53.75→68.75% / 42.5→62.5% / 31.25→43.75% against a 90 / 75-80 / 60-70 target.
What changed
per-preset floors + dominance ceiling, a 70/70 card effectiveness lint, keyword-atlas gate marks, RUPTURE/REAP-ALL burst caps scaled to max(floor, 0.25 × enemy maxHealth), and the CONCEDE/CAPITULATE win-path scaling item (boss rung-regrowth capped at 2× natural). Oratory/standstill still hit the 1.00 ceiling late — reclassified as a /deck-tuning BACKFIRE-tuning item, not an engine threshold; DoT persistence-by-stack and the theme pressure valve remain open.
Why it changed
continuation of the challenge-gradient retune — the burst-cap formula alone erased the Penitent late-game regression (0.02→0.40).
commits 43088f6f · 60b1e26b · 8f646629 · 8d853dcb · 679ba168 · 9eaaf8d7 · 63cfb3ba
What changed
three PRs retired the last vestiges of the pre-Hazard-Pattern "skill" combat system: unified skill→card across the combat system and deleted the dead src/Skills/ ability system (PR #48); dropped the save-migration chain and the physicalSkill/mentalSkill/ emotionalSkill derived-stat axis (PR #49); deleted the dead CombatState.log befriend machinery — requiredCardUse was authored on zero enemies, requiredStances on nine but never evaluated true under the Hazard-Pattern shim's always-empty log — and renamed the residual skillId to cardId/signatureId by what it actually holds (PR #50). A fourth PR (ci-autofix) caught and fixed a stale learningRequirement.level reference in card-editor within three minutes of the red run. plan/cleanup/legacy-skill-removal.md tracks WI-2 through WI-6 as the remaining backlog (effect content, data enums, one-offs, comments, docs); the only surviving "skill" is now SignatureSkill.
Why it changed
the rename had been chased piecemeal for weeks; this was the first pass with a written definition of done (grep -rioE "\bskills?\b" returns zero non-SignatureSkill hits).
commits 5d5fcca7 · 911dd3c4 · 4e9d9d01 · ae48ae2a · ffab2cc5 · 8d6baba0 · 44c9bb97 · f3aa8900 · 9fa7f670 · f5a96468 · 55ee82a3 · 51479b05 · d0c7fdd9
What changed
march cron dropped from hourly to every 6h, preview-build from 2×/day to 1×/day, and nine weekly tuning/report crons disabled (workflow_dispatch still available) until the next billing cycle.
Why it changed
hit 90% of the monthly Actions-minute quota within days of turning the crons on.
commit 47259bec
What changed
tagged cards with their originating deck-preset theme (for the catalog export), fixed the FREE-line enchantment/disenchant effects, and corrected "signature" prose to HALLMARK terminology across agents, commands, and the card-editor face.
Why it changed
small follow-through on the Themed Deck Library v3 / enchantment-redesign work from the prior digest.
What changed
reorganized the devlog into a cleaner structure and added a devlog/tuning-lab/ section holding the Battle Lab round 2 HTML artifacts.
Why it changed
the round-2 conclusions previously lived only in an out-of-repo HTML artifact; this gives them a permanent, linkable home next to the atlas gate-mark update (win-path scaling item 7).
✠ The marginalia
| Time (UTC) | Verb | Outcome |
|---|---|---|
| 2026-07-08 11:12 | merge (owner, PR #41) | thematic rebalance of all 10 preset decks, two verify-caught regressions fixed pre-merge |
| 2026-07-08 11:17 → 15:07 | owner (direct, Battle Lab round 2) | win-path-scaling plan authored; items 4, 6, 7, 2, 1 shipped in sequence |
| 2026-07-08 11:58 → 12:07 | march | critique pass 7 dispatched in-tick |
| 2026-07-08 12:05 | critique pass 7 | 0 findings |
| 2026-07-08 14:42 → 14:53 | march | phase 9 shipped |
| 2026-07-08 14:55 → 15:07 | owner (direct) | tuning-lab devlog pages added |
| 2026-07-08 16:51 → 17:03 | march | critique pass 8 — 0 findings, 4th consecutive playwright block |
| 2026-07-08 18:36 → 20:27 | owner (direct) | card theme tagging, card edits, enchantment fixes, HALLMARK terminology pass |
| 2026-07-08 18:49 → 19:27 | march | phase 10 shipped (cross-screen integration harness) |
| 2026-07-08 20:38 → 20:49 | march | critique pass 9 — 0 findings, 5th consecutive playwright block |
| 2026-07-08 22:00 → 23:53 | owner (direct, PRs #48/#49) | skill→card purge waves 1-2: combat system unified, save-migration + Skill stat axis dropped |
| 2026-07-08 22:18 → 22:42 | march | phase 11 shipped (rest tutorial) |
| 2026-07-08 22:51 | ci (owner, direct) | Actions cron schedules throttled — 90% of monthly minute quota hit |
| 2026-07-08 23:31 → 23:49 | march | phase 12 shipped (doc-sync pass) |
| 2026-07-09 00:15 → 00:18 | ci-autofix (PR #48 branch) | verify-mechanics/verify-card-editor red — stale learningRequirement.level caught and fixed in 3 minutes |
| 2026-07-09 00:22 → 01:00 | owner (direct, PR #50) | WI-1 — dead CombatState.log befriend machinery retired; residual skillId renamed |
| 2026-07-09 01:25 → 01:31 | owner (direct) | phases 18-21 briefs added — equipment/signature refactor epic queued |
| 2026-07-09 10:00 → 10:10 | march | no-op tick |
| 2026-07-09 11:34 (this run) | night | digest — breadth check needed the same manual Playwright browser install as every prior night run |
Build-plan queue (plan/steps/01_build_plan.md): 14/22 shipped (phases 0, 14, 1-12) — four shipped by the loop since yesterday (9, 10, 11, 12); phases 18-21 (equipment/signature refactor epic) newly queued by the owner, bringing pending to 8 (13, 15-21). plan/AUDIT.md 7 pending (unchanged). plan/CRITIQUE.md 3 pending (same recurring rows, now at critique pass 9 on the playwright block). plan/PHASE_CANDIDATES.md 9 pending (/expand still hasn't run since the 2026-07-03 seed). No open triage:needs-user or loop:do issues. Deploy gate: no gated workflow for HEAD yet (grace window) — nothing to check. Breadth check (npm run e2e:minigames, browser installed by hand): hazard, combat, gathering, encounter-routing, exploration-roundtrip — ALL PASS.
None filed by digest this pass (in scope: loop-gate mistuning only — game-balance tuning routes through /expand or the domain tuning skills). The Actions-minute throttle above is an operational capacity change already applied by the owner directly, not a gate constant this loop can propose against.
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.