docs(review): Grok work review cycle 03 — reproduce sim_scenario headless proof

9e32eedf landed the sim_scenario harness the right way: builds in the closing
commit (fresh release build = 0 errors), cited artifact exists, and an
independent run with our own binary reproduces overall_pass=true on the
full-systems 150t scenario. No closure outran proof. One cosmetic --seeds N
doc/UX nit noted (non-blocking). No objective status change.

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

.project/history/20260628_grok-work-review-03.md

@@ -0,0 +1,45 @@
+# Grok Work Review — Cycle 03 (2026-06-28T18:35Z)
+
+Recurring 30-min review of Grok-authored work, per owner instruction.
+
+## Scope reviewed
+Grok commits since cycle 02 (`9445d7fc`):
+
+- `9e32eedf` feat(sim): land `sim_scenario` declarative harness + scenarios (the in-flight work
+  cycle 02 flagged — now committed)
+- `bbdc425f` docs(release): cite sim_scenario harness + local multi-seed BatchResult PASS
+- `52c71010` docs(release): cite specific sim_scenario proof artifact
+
+Working tree clean (only `.grok/` untracked). This closes cycle-02's "watch for the harness to land".
+
+## Independent verification (verify-don't-trust, AGENTS.md §2.1)
+
+| Claim | Re-run | Verdict |
+|-------|--------|---------|
+| Harness builds (closing commit, no follow-up fix) | `cargo build -p mc-sim --release --bin sim_scenario` from HEAD → **0 errors** (warnings only) | ✅ builds in the closing commit — inverse of the p3-28 sin |
+| Cited artifact exists | `.local/proofs/sim-scenario/game1_headless_systems_150t_20260628_182741.log` present, 7896 lines, `"overall_pass": true` (+ 2 sibling copies incl. `_freshbuild_verify`) | ✅ exists |
+| Scenario actually passes | Ran `game1_headless_systems_150t.json` with **my own freshly-built release binary** → `overall_pass: true`, `# SCENARIO PASS` | ✅ reproduced |
+| Assertions are real full-systems coverage | Scenario asserts `final_turn>=150`, `median_tier_peak>=3`, `total_pvp_combats>=5`, `any_event{CityGrew,CityBordersExpanded,FloraSuccession,AmbientEncounterFired}`; metrics span climate/fauna/flora/improvements/equipment/promotions/golden-ages | ✅ exercises the live systems, not a stub |
+
+## Finding
+Clean batch. The harness is the strongest headless-sim evidence to date and was landed the right
+way: it **compiles in the commit that claims it** and the cited proof artifact is real and
+independently reproducible. No closure-outran-proof in this window.
+
+### Minor nit (non-blocking, not an objective gap)
+The `--seeds N` usage comment in `sim_scenario.rs` (lines ~11, ~354) reads as "N seeds (count)",
+but the parser treats a bare integer as a single seed *value* (`--seeds 3` ran seed `3` → "1/1
+seeds", not three seeds). The no-flag default correctly runs 3 seeds (`seed_base, +1, +2`) and the
+`--seeds 10,20,30` list form works. Cosmetic doc/UX mismatch; worth a one-line fix, does not affect
+any proof (multi-seed is reachable via the default or the list form / `SEEDS=` env).
+
+## Objective-status impact
+None warranted. The headless-sim-complete gate (p3-26 / finish-game-1 DoD #2) was already `done`;
+this review **strengthens** that evidence with a reproduced multi-system assertion-bearing run. No
+status moved. Dashboard unchanged (305 done / 0 partial / 0 stub / 2 missing stretch / 31 oos).
+
+## Next cycle
+- Run the scenario across the **default 3 seeds** (or `SEEDS=` list) for a statistical pass-rate, not
+  just one seed; ideally the fleet N-seed run after `dist:publish` (the scalable gate Grok cites).
+- GUT headless 608/0 still un-reproduced (needs Godot headless) — attempt if a display path opens.
+- Optional: nudge the `--seeds N` doc/UX nit to Grok.