docs(review): Grok work review cycle 02 — close mc-player-api gate (142 lib/0), assess in-flight sim_scenario harness

No new committed Grok work since cycle 01. In-flight uncommitted sim_scenario
runner compiles clean (0 err); design sound (Rail-1/Rail-2 aligned), correctly
not yet claimed done. mc-player-api reproduced 142 lib + 42 integ = 184/0,
matching eca713bf. No objective status changes warranted.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.project/history/20260628_grok-work-review-02.md

@@ -0,0 +1,43 @@
+# Grok Work Review — Cycle 02 (2026-06-28T17:54Z)
+
+Recurring 30-min review of Grok-authored work, per owner instruction.
+
+## Scope reviewed
+- **New committed Grok work since cycle 01 (`93d7fd16`):** none.
+- **In-flight (uncommitted) Grok work:** a new `mc-sim` declarative scenario-runner —
+  `src/simulator/crates/mc-sim/src/bin/sim_scenario.rs` (+ `[[bin]]` decl), scenario data under
+  `public/games/age-of-dwarves/data/sim-scenarios/` (`smoke_duel_30t.json`,
+  `game1_headless_systems_150t.json`, `combat/`), and doc wiring (AGENTS.md §2.1 now names
+  `sim_scenario` as the preferred statistical sim-proof primitive; `scripts/run/dist.sh`,
+  `simulator-infra.md`, `agents-task-map.md`, `finish-game-1/SKILL.md`).
+
+## Independent verification
+
+| Check | Result | Notes |
+|-------|--------|-------|
+| `cargo check -p mc-sim --bin sim_scenario` | ✅ 0 errors (1 doc warn) | Grok's in-flight harness compiles |
+| `cargo test -p mc-player-api` (cycle-01 open gate) | ✅ 142 lib + 42 integ = **184 / 0 failed** | lib count **142** matches Grok's `eca713bf` claim exactly |
+
+### Note on a measurement race (reviewer error, not Grok's)
+A first hurried `mc-player-api` run appeared to show only "1 passed". That was a wait-loop that
+exited on the first (fast lib) result line before the integration binaries ran — re-running cleanly
+gave the full 184/0. Recorded here for honesty; no defect in Grok's work.
+
+## Assessment of the in-flight scenario harness
+Design is sound: declarative JSON scenario (map/players/rules/metrics/assertions), seeded multi-run,
+machine-readable BatchResult, intended for horizontal fan-out on the DO fleet after `dist:publish`.
+Aligns with Rail-1 (drives `mc-turn::TurnProcessor` + worldsim pre-pass + mc-ai personalities; no
+new sim logic in GDScript) and Rail-2 (scenarios are JSON content). It is **not yet committed** — so
+per AGENTS.md §2 it is correctly *not* claimed done anywhere; no objective should flip on it until
+Grok commits + cites a fleet run artifact.
+
+## Objective-status impact
+None. No committed status changes since cycle 01; in-flight work is correctly un-closed. Dashboard
+unchanged (305 done / 0 partial / 0 stub / 2 missing stretch / 31 oos). The cycle-01 open
+verification gate (mc-player-api) is now **reproduced green**.
+
+## Next cycle
+Watch for Grok to commit the `sim_scenario` harness — when it lands, verify it builds in the closing
+commit (not a follow-up fix), run a scenario locally, and confirm the cited fleet artifact exists.
+GUT headless 608/0 still un-reproduced (needs Godot headless) — attempt next cycle if a display path
+is available.