diff --git a/scripts/impasse_report.py b/scripts/impasse_report.py
index 1dde22e..ba6b822 100644
--- a/scripts/impasse_report.py
+++ b/scripts/impasse_report.py
@@ -178,11 +178,22 @@ def render(run: dict) -> str:
     by = {"accepted": 0, "rejected": 0, "resolved": 0, "deadlocked": 0, "withdrawn": 0}
     for it in items.values():
         by[it.get("state")] = by.get(it.get("state"), 0) + 1
+    # An operator ruling counts as an escalation regardless of channel (SKILL.md): a resolved item
+    # carrying an escalation object was decided BY the operator, not settled between the two models.
+    # Credit it separately so the tally can't under-report the operator's own involvement (issue #5) —
+    # `pending` is the still-deadlocked work, `resolved_shown` the truly model-settled resolutions.
+    decided_by_you = sum(1 for it in items.values()
+                         if it.get("state") == "resolved" and isinstance(it.get("escalation"), dict))
+    pending = by["deadlocked"]
+    resolved_shown = by["resolved"] - decided_by_you
     out.append("")
-    out.append(
-        f"📊 Decisions: {n} finding(s) raised → ✅ {by['resolved']} resolved · 🤝 {by['accepted']} accepted · "
-        f"❌ {by['rejected']} rejected · ⚖️ {by['deadlocked']} escalated to you"
+    tally = (
+        f"📊 Decisions: {n} finding(s) raised → ✅ {resolved_shown} resolved · 🤝 {by['accepted']} accepted · "
+        f"❌ {by['rejected']} rejected · ⚖️ {pending} escalated to you"
     )
+    if decided_by_you:
+        tally += f" · 🧑‍⚖️ {decided_by_you} decided by you"
+    out.append(tally)
     if rec.get("failure"):
         out.append(f"⚠️ Failure: {rec['failure'].get('code')} — {rec['failure'].get('message')}")
     out.append("─" * 78)
@@ -192,9 +203,15 @@ def render(run: dict) -> str:
         out += _render_finding(f, items.get(f.get("id")))
         out.append("─" * 78)
 
-    esc = by["deadlocked"]
-    if esc:
-        out.append(f"⚖️  {esc} decision(s) need you; the rest the models settled between themselves.")
+    if pending:
+        if decided_by_you:
+            out.append(f"⚖️  {pending} decision(s) need you; you decided {decided_by_you}; "
+                       "the rest the models settled between themselves.")
+        else:
+            out.append(f"⚖️  {pending} decision(s) need you; the rest the models settled between themselves.")
+    elif decided_by_you:
+        out.append(f"✅  Models settled {len(items) - decided_by_you}; you decided {decided_by_you}. "
+                   "Nothing is waiting on you.")
     elif items:
         out.append(f"✅  Nothing needed you — the models settled all {len(items)} between themselves.")
     else:
diff --git a/tests/test_helpers.py b/tests/test_helpers.py
index a75ff40..3c80712 100644
--- a/tests/test_helpers.py
+++ b/tests/test_helpers.py
@@ -1409,6 +1409,39 @@ def _spy_cwd(argv, **kw):
     check("Question for you" in out and "decision(s) need you" in out, "report: shows the escalated question")
     check(any(r["run_id"] == drid for r in lib.list_runs()), "run record: listed by list_runs")
 
+    # --- report credits operator-decided items (issue #5): a resolved item carrying an escalation
+    # object was decided BY the operator, not settled between the models — the tally + footer must say so ---
+    _b_run1 = {"reviewer_response": {}, "reconciliation_result": {"review_id": "b-1", "items": [
+        {"finding_id": "F1", "state": "resolved", "resolution": "Operator chose repair A over B.",
+         "escalation": {"dispute_kind": "value_or_priority_tradeoff",
+                        "stop_reason": "operator_authority_required", "operator_question": "A or B?"}},
+        {"finding_id": "F2", "state": "accepted"}]}}
+    _b_out1 = report.render(_b_run1)
+    check("decided by you" in _b_out1 and "you decided" in _b_out1
+          and "Nothing is waiting on you" in _b_out1 and "Nothing needed you" not in _b_out1,
+          "report: operator-decided (resolved + escalation) item is credited, not counted as autonomous")
+    _b_run2 = {"reviewer_response": {}, "reconciliation_result": {"review_id": "b-2", "items": [
+        {"finding_id": "F1", "state": "resolved", "resolution": "host fix"},
+        {"finding_id": "F2", "state": "accepted"}]}}
+    _b_out2 = report.render(_b_run2)
+    check("Nothing needed you — the models settled all 2 between themselves." in _b_out2
+          and "decided by you" not in _b_out2,
+          "report: genuinely-autonomous run still says 'Nothing needed you'")
+    _b_run3 = {"reviewer_response": {}, "reconciliation_result": {"review_id": "b-3", "items": [
+        {"finding_id": "F1", "state": "deadlocked",
+         "escalation": {"dispute_kind": "x", "stop_reason": "y", "operator_question": "q1?"}},
+        {"finding_id": "F2", "state": "resolved", "resolution": "op ruling",
+         "escalation": {"dispute_kind": "z", "stop_reason": "w", "operator_question": "q2?"}}]}}
+    _b_out3 = report.render(_b_run3)
+    # Mixed case: pending deadlock AND an operator-decided item. The footer stays on the pending
+    # branch, but must CREDIT the operator ("you decided 1") — not attribute their ruling to the
+    # models. The footer is the last rendered line; "you decided" appears only there (not the tally).
+    _b_foot3 = _b_out3.splitlines()[-1]
+    check("decision(s) need you" in _b_out3 and "escalated to you" in _b_out3
+          and "decided by you" in _b_out3
+          and "you decided 1" in _b_foot3 and "decision(s) need you" in _b_foot3,
+          "report: a deadlock still takes footer precedence over an operator-decided item")
+
     # --- lifetime recap: aggregate value across reconciled runs (isolated config dir) ---
     recap_dir = tempfile.mkdtemp(prefix="impasse-recap-")
     _prev_cfg = os.environ["IMPASSE_CONFIG_DIR"]