feat: online verification, node removals, and knowledge browser

Adds claim-bound online evidence records and citation verification scripts,
node removal transport/adapter, retrieval benchmark and owner-review tooling,
GraphCanvas/SourceDetail knowledge browser UI, and the ingest handoff helper.
Bumps the agents/gewu pack to the online-verification branch.

tests/test_run_four_file_automation.py

@@ -0,0 +1,77 @@
+from scripts.run_four_file_automation import gate_summary
+
+
+def test_gate_summary_does_not_qualify_when_protected_and_answer_gates_remain():
+    report = gate_summary(
+        source_checks=[{"status": "pass"}],
+        review_ledger={"summary": {"evidence_bindings_invalid": 0, "claims": 2, "items": 1}},
+        retrieval_score={
+            "question_count": 24,
+            "mean_conditional_recall_at_5": 0.9,
+            "conditional_hit_rate": 1.0,
+            "pagination_blocked_queries": 0,
+        },
+        answer_reports={
+            "answers": {"status": "invalid", "checked": 1, "issues": [{"code": "BAD"}]}
+        },
+        support_judgments={"strict_group_support_fraction": 1.0},
+        answer_scores={"50k": {"content_points_before_citation_penalty": 48, "max_points": 48}},
+        behavioral_results={
+            "exact_replay": {"passed": True},
+            "quote": {"status": "invalid"},
+            "offset": {"status": "invalid"},
+            "extraction": {"status": "invalid"},
+            "revision": {"status": "invalid"},
+            "meaning": {"status": "eligible"},
+            "unauthorized_get_node": {"denied": True},
+            "unauthorized_search": {"denied": True},
+        },
+    )
+
+    assert report["qualification"] == "not_qualified"
+    assert report["gates"]["retrieval_tuning"]["status"] == "pass"
+    assert report["gates"]["answer_citations"]["status"] == "fail"
+    assert report["gates"]["owner_protected_decisions"]["status"] == "pending_owner"
+    assert report["gates"]["behavioral_suite"]["status"] == "partial"
+    assert report["gates"]["semantic_support_review"]["status"] == "gap_demonstrated_not_judged"
+    assert "reserved_questions" in report["remaining_gates"]
+
+
+def test_gate_summary_marks_retrieval_failure_when_pagination_is_blocked():
+    report = gate_summary(
+        source_checks=[{"status": "pass"}],
+        review_ledger={"summary": {"evidence_bindings_invalid": 0, "claims": 0, "items": 0}},
+        retrieval_score={
+            "question_count": 24,
+            "mean_conditional_recall_at_5": 0.9,
+            "conditional_hit_rate": 1.0,
+            "pagination_blocked_queries": 1,
+        },
+        answer_reports={"answers": {"status": "valid", "checked": 0, "issues": []}},
+        support_judgments={"strict_group_support_fraction": 1.0},
+        answer_scores={"50k": {"content_points_before_citation_penalty": 48, "max_points": 48}},
+    )
+
+    assert report["gates"]["retrieval_tuning"]["status"] == "fail"
+    assert report["qualification"] == "not_qualified"
+
+
+def test_semantic_judge_completion_remains_advisory_until_candidate_actions_finish():
+    report = gate_summary(
+        source_checks=[{"status": "pass"}],
+        review_ledger={"summary": {"evidence_bindings_invalid": 0, "claims": 1, "items": 1}},
+        retrieval_score={"question_count": 24, "mean_conditional_recall_at_5": 1.0},
+        answer_reports={"answers": {"status": "valid", "checked": 0, "issues": []}},
+        support_judgments={"strict_group_support_fraction": 1.0},
+        answer_scores={"50k": {"content_points_before_citation_penalty": 48, "max_points": 48}},
+        semantic_results={
+            "claim_count": 1,
+            "outcomes": {"advisory_support_consensus": 1},
+        },
+    )
+
+    assert (
+        report["gates"]["semantic_support_review"]["status"]
+        == "advisory_complete_candidate_actions_pending"
+    )
+    assert report["qualification"] == "not_qualified"