diff --git a/public/games/age-of-dwarves/data/sim-scenarios/fullgame/clan_fairness_band.json b/public/games/age-of-dwarves/data/sim-scenarios/fullgame/clan_fairness_band.json
index d3a47bc8..52cdf6a8 100644
--- a/public/games/age-of-dwarves/data/sim-scenarios/fullgame/clan_fairness_band.json
+++ b/public/games/age-of-dwarves/data/sim-scenarios/fullgame/clan_fairness_band.json
@@ -2,7 +2,8 @@
   "id": "clan_fairness_band",
   "kind": "fullgame",
   "version": 1,
-  "description": "Balance gate: round-robin all 5+ clan personalities across many seeds; no single personality may exceed the win-rate ceiling. A statistical scenario meant for the DO fleet (needs many games). Threshold is calibrated, not aspirational.",
+  "gating": false,
+  "description": "NON-GATING (owner decision 2026-06-30). Measures SCRIPTED clan-personality balance: round-robin all 6 personalities across 50 seeds. Current reality — tech_rusher ~46%, turtle 40%, militarist 14%, and boom/expansionist/merchant 0% — a real imbalance in the SCRIPTED AI. The project's answer is the TRAINED/learned controllers (which perform much better), NOT scripted-personality rebalancing, so this does not gate the release baseline; it stays as a measurement. Fix path: train learned controllers toward the 6 clan types (docs/ai-roadmap.md). The 0.4 ceiling (a pre-calibration placeholder) is intentionally left untuned so the gap stays visible.",
   "map": { "size": 32, "evolution_ticks": 12000, "seed_base": 10000 },
   "players": [
     { "personality": "militarist" }, { "personality": "boom" },
diff --git a/scripts/green-pass.sh b/scripts/green-pass.sh
new file mode 100755
index 00000000..52a561a3
--- /dev/null
+++ b/scripts/green-pass.sh
@@ -0,0 +1,70 @@
+#!/usr/bin/env bash
+# Hardened-baseline gate: the full horizontal green pass for the simulator.
+#
+#   1. cargo nextest run --workspace   — every crate's unit + integration tests
+#   2. all sim scenarios               — combat set-pieces + fullgame invariants,
+#                                        run through the REAL sim_scenario resolver
+#
+# Gating: a scenario JSON with "gating": false is run and REPORTED but does not
+# fail the gate (e.g. clan_fairness_band measures scripted-personality balance,
+# which is intentionally superseded by trained/learned controllers — see
+# docs/ai-roadmap.md). Every other scenario, and any workspace test failure,
+# fails the gate.
+#
+# Designed to run ON a fleet worker (no local Rust toolchain on plum):
+#   ./run dist:up 1 && rsync this repo state up && ssh worker 'bash scripts/green-pass.sh'
+# Exit 0 only when the baseline is fully green.
+set -uo pipefail
+
+REPO="${MC_REPO:-$HOME/Code/@projects/@magic-civilization}"
+cd "$REPO" || { echo "no repo at $REPO" >&2; exit 2; }
+. ~/.cargo/env 2>/dev/null || true
+SCN_DIR="public/games/age-of-dwarves/data/sim-scenarios"
+OUT="${MC_GREENPASS_OUT:-/tmp/green-pass}"
+rm -rf "$OUT"; mkdir -p "$OUT"
+echo "HEAD: $(git log --oneline -1 2>/dev/null)"
+
+# ── [1/3] workspace tests ────────────────────────────────────────────────────
+echo "########## [1/3] cargo nextest run --workspace ##########"
+( cd "$REPO/src/simulator" && cargo nextest run --workspace --no-fail-fast ) > "$OUT/tests.log" 2>&1
+TESTS_RC=$?
+tail -4 "$OUT/tests.log"
+echo "tests rc=$TESTS_RC"
+
+# ── [2/3] build the scenario runner ──────────────────────────────────────────
+echo "########## [2/3] build sim_scenario ##########"
+( cd "$REPO/src/simulator" && cargo build --release -p mc-sim --bin sim_scenario ) > "$OUT/simbuild.log" 2>&1
+SIMBIN="$(find "$REPO" -type f -name sim_scenario -path '*release*' 2>/dev/null | head -1)"
+echo "sim_scenario: ${SIMBIN:-NOT FOUND}"
+
+# ── [3/3] run scenarios (gating-aware) ───────────────────────────────────────
+echo "########## [3/3] scenarios ##########"
+gate_pass=0; gate_fail=0; nongate_fail=0; failed_gating=""; failed_nongating=""
+if [ -x "$SIMBIN" ]; then
+  for f in "$SCN_DIR"/combat/*.json "$SCN_DIR"/fullgame/*.json; do
+    [ -e "$f" ] || continue
+    name="$(basename "$f" .json)"
+    gating="$(python3 -c "import json,sys;print(json.load(open('$f')).get('gating',True))" 2>/dev/null)"
+    if "$SIMBIN" "$f" > "$OUT/scn_$name.json" 2>"$OUT/scn_$name.err"; then
+      echo "PASS  $name"; gate_pass=$((gate_pass+1))
+    elif [ "$gating" = "False" ]; then
+      echo "KNOWN $name  (non-gating — does not fail the baseline)"; nongate_fail=$((nongate_fail+1)); failed_nongating="$failed_nongating $name"
+    else
+      echo "FAIL  $name"; gate_fail=$((gate_fail+1)); failed_gating="$failed_gating $name"
+    fi
+  done
+else
+  echo "sim_scenario not built — gate fails"; gate_fail=1
+fi
+
+echo "=================== SUMMARY ==================="
+echo "workspace tests rc=$TESTS_RC"
+echo "scenarios: $gate_pass gating-pass, $gate_fail gating-FAIL, $nongate_fail non-gating-known-fail"
+[ -n "$failed_gating" ]    && echo "GATING_FAILURES:$failed_gating"
+[ -n "$failed_nongating" ] && echo "KNOWN_NONGATING:$failed_nongating"
+
+if [ "$TESTS_RC" -eq 0 ] && [ "$gate_fail" -eq 0 ]; then
+  echo "GREEN_BASELINE: OK"; exit 0
+else
+  echo "GREEN_BASELINE: FAILED"; exit 1
+fi