opencode-skills-collection 4.0.69 → 4.0.70

package/bundled-skills/fact-check-x-unified/tests/smoke_test.py

@@ -0,0 +1,740 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import json
+import os
+import shutil
+import subprocess
+import sys
+import tempfile
+import zipfile
+from pathlib import Path
+from unittest.mock import patch
+
+
+ROOT = Path(__file__).resolve().parents[1]
+SKILLS = ROOT.parent
+if (ROOT / "modules").exists():
+    SKILLS = ROOT / "modules"
+COMPARE_FIXTURES = SKILLS / "fact-check-x-knowledge-compare" / "tests" / "fixtures"
+AUTHORITY_FIXTURES = SKILLS / "fact-check-x-authoritative-verify" / "tests" / "fixtures"
+
+
+def run(arguments: list[str], environment: dict[str, str] | None = None) -> dict:
+    process = subprocess.run(arguments, text=True, capture_output=True, check=False, env=environment)
+    if process.returncode:
+        raise AssertionError(process.stdout or process.stderr)
+    payload = json.loads([line for line in process.stdout.splitlines() if line.strip()][-1])
+    checkpoint = payload.get("checkpoint") or {}
+    acknowledgement = checkpoint.get("acknowledgement") or {}
+    if checkpoint.get("status") == "awaiting_user" and acknowledgement:
+        acknowledged = subprocess.run(
+            [sys.executable, str(ROOT / "scripts" / "fact_check_x.py"),
+             "acknowledge-stage", "--run-dir", str(Path(checkpoint["path"]).parent),
+             "--stage", checkpoint["stage"], "--token", acknowledgement["token"],
+             "--decision", "continue"],
+            text=True, capture_output=True, check=False, env=environment,
+        )
+        if acknowledged.returncode:
+            raise AssertionError(acknowledged.stdout or acknowledged.stderr)
+    return payload
+
+
+def run_failed(arguments: list[str], environment: dict[str, str] | None = None) -> dict:
+    process = subprocess.run(arguments, text=True, capture_output=True, check=False, env=environment)
+    if process.returncode == 0:
+        raise AssertionError("命令应以非零状态结束")
+    return json.loads([line for line in process.stdout.splitlines() if line.strip()][-1])
+
+
+def command(*arguments: str) -> list[str]:
+    return [sys.executable, str(ROOT / "scripts" / "fact_check_x.py"), *arguments]
+
+
+def main() -> int:
+    with tempfile.TemporaryDirectory(prefix="fact-check-x-unified-") as temp:
+        sys.path.insert(0, str(ROOT / "scripts"))
+        import fact_check_x
+        from fact_check_x import (
+            collect_technical_notices,
+            merge_verification,
+            normalize_report_navigation,
+        )
+        assert fact_check_x.file_uri_for_path(
+            r"C:\WorkBuddy\fact-check-x\runs\sample run\02-comparison-report.html"
+        ) == "file:///C:/WorkBuddy/fact-check-x/runs/sample%20run/02-comparison-report.html"
+        assert fact_check_x.normalize_platform_path_text(
+            "/c/WorkBuddy/fact-check-x/runs/sample run/02-comparison-report.html",
+            "nt",
+        ) == "C:/WorkBuddy/fact-check-x/runs/sample run/02-comparison-report.html"
+        assert fact_check_x.file_uri_for_path(
+            "/c/WorkBuddy/fact-check-x/runs/sample run/02-comparison-report.html",
+            "nt",
+        ) == "file:///C:/WorkBuddy/fact-check-x/runs/sample%20run/02-comparison-report.html"
+
+        def assert_deliverable(item: dict, label: str, path: Path) -> None:
+            uri = path.resolve().as_uri()
+            assert item["label"] == label
+            assert item["path"] == str(path.resolve())
+            assert item["fileUri"] == uri
+            assert item["markdownLink"] == f"[打开{label}](<{uri}>)"
+
+        notice_run = Path(temp) / "technical-notice"
+        (notice_run / "capture").mkdir(parents=True)
+        (notice_run / "capture" / "results.json").write_text(
+            json.dumps({
+                "platforms": [{
+                    "platform": "doubao",
+                    "label": "豆包",
+                    "references": [
+                        {"sourceAcquisitionStatus": "blocked"},
+                        {"contentAcquisition": "failed"},
+                    ],
+                }],
+            }, ensure_ascii=False),
+            encoding="utf-8",
+        )
+        notices = collect_technical_notices(notice_run)
+        assert len(notices) == 1 and "豆包 2 条" in notices[0]
+        notice_html = normalize_report_navigation(
+            b"<html><head><style></style></head><body><main>test</main></body></html>",
+            "01-capture-report.html",
+            notices,
+        ).decode("utf-8")
+        assert 'data-fcx-run-notice="1"' in notice_html
+        assert "受影响且无法核验的主张按“疑似误导”呈现" in notice_html
+
+        merge_results = Path(temp) / "merge-results"
+        merge_results.mkdir()
+        (merge_results / "K1.json").write_text(json.dumps({
+            "schemaVersion": "fact-check-x/authority-result@1",
+            "requestId": "K1",
+            "searchMode": "trusted_search",
+            "requestCount": 1,
+            "authoritativeFinding": "直接问题权威结论。",
+            "claims": {
+                "doubao": {"covered": True, "claim": "直接问题平台主张。"},
+            },
+            "verdicts": {
+                "doubao": {"verdict": "contradicted"},
+            },
+            "evidenceGaps": [],
+        }), encoding="utf-8")
+        (merge_results / "K2.json").write_text(json.dumps({
+            "schemaVersion": "fact-check-x/authority-result@1",
+            "requestId": "K2",
+            "searchMode": "trusted_search",
+            "requestCount": 1,
+            "authoritativeFinding": "补充参考权威结论。",
+            "claims": {
+                "doubao": {"covered": True, "claim": "补充参考平台主张。"},
+            },
+            "verdicts": {
+                "doubao": {"verdict": "supported"},
+            },
+            "evidenceGaps": [],
+        }), encoding="utf-8")
+        merged = merge_verification({
+            "question": "测试",
+            "platforms": [{"platform": "doubao", "label": "豆包"}],
+            "knowledgePoints": [{
+                "id": "K1",
+                "description": "直接问题",
+                "role": "direct",
+                "trustedAnchor": {"eligible": False},
+            }, {
+                "id": "K2",
+                "description": "补充参考",
+                "role": "reference",
+                "trustedAnchor": {"eligible": False},
+            }],
+            "analysisGaps": [{
+                "stage": "comparison",
+                "knowledgePointId": "K1",
+                "platform": "doubao",
+                "reason": "原回答引用不足",
+            }],
+        }, merge_results)
+        assert merged["status"] == "completed"
+        assert merged["evidenceGaps"] == []
+        assert merged["finalAnswer"]["knowledgePointIds"] == ["K1"]
+        assert merged["supplementalFindings"]["knowledgePointIds"] == ["K2"]
+        assert merged["finalAnswer"]["answer"] == "直接问题权威结论。"
+        assert "补充参考权威结论" not in merged["finalAnswer"]["answer"]
+
+        run_dir = Path(temp) / "run"
+        results = COMPARE_FIXTURES / "results.json"
+        keyless_environment = {
+            **os.environ,
+            "TRUSTED_SEARCH_KEY": "",
+            "FACT_CHECK_X_TRUSTED_SEARCH_KEY_FILE": str(
+                Path(temp) / "missing-trusted-search-key"
+            ),
+        }
+        located = run(command("locate"))
+        assert set(located["skills"]) == {"collector", "comparison", "authority"}
+        capture_stage = run(command("prepare-comparison", "--results", str(results), "--run-dir", str(run_dir)))
+        assert capture_stage["stage"] == "capture_completed"
+        assert capture_stage["analysisExecution"]["mode"] == "single_pass"
+        assert capture_stage["analysisExecution"]["maxToolCallsAfterStageAcknowledgement"] == 3
+        assert len(capture_stage["analysisExecution"]["instructions"]) == 3
+        assert capture_stage["artifacts"]["answerReferenceReport"] == str((run_dir / "capture" / "report.html").resolve())
+        assert capture_stage["deliverables"][0]["path"] == str((run_dir / "01-capture-report.html").resolve())
+        assert_deliverable(
+            capture_stage["deliverables"][0],
+            "各方答案汇总",
+            run_dir / "01-capture-report.html",
+        )
+        assert capture_stage["checkpoint"]["mustPresentBeforeNextStage"] is True
+        assert capture_stage["checkpoint"]["path"] == str(
+            (run_dir / "01-capture-report.html").resolve()
+        )
+        assert "打开各方答案汇总" in capture_stage["checkpoint"]["message"]
+        assert (run_dir / "01-capture-report.html").exists()
+        assert len(capture_stage["platforms"]) == 2
+        comparison_stage = run(command("complete-comparison", "--results", str(results), "--analysis", str(COMPARE_FIXTURES / "comparison-analysis.json"), "--run-dir", str(run_dir)))
+        assert comparison_stage["stage"] == "comparison_completed"
+        assert comparison_stage["knowledgePointCount"] == 1
+        assert comparison_stage["artifacts"]["comparisonAnalysis"] == str(
+            (run_dir / "comparison-analysis.json").resolve()
+        )
+        assert (run_dir / "comparison-analysis.json").is_file()
+        assert comparison_stage["artifacts"]["comparisonReport"] == str((run_dir / "comparison.html").resolve())
+        assert comparison_stage["deliverables"][0]["path"] == str((run_dir / "02-comparison-report.html").resolve())
+        assert_deliverable(
+            comparison_stage["deliverables"][0],
+            "各方答案聚合（未核验）",
+            run_dir / "02-comparison-report.html",
+        )
+        assert comparison_stage["checkpoint"]["mustPresentBeforeNextStage"] is True
+        assert comparison_stage["checkpoint"]["path"] == str(
+            (run_dir / "02-comparison-report.html").resolve()
+        )
+        overwide_analysis = json.loads(
+            (COMPARE_FIXTURES / "comparison-analysis.json").read_text(encoding="utf-8")
+        )
+        overwide_analysis["knowledgePoints"][0]["claims"]["dknowc-chat"]["claim"] = (
+            "每人每月最高提取1400元，并另有最高300万元研发资助"
+        )
+        overwide_path = Path(temp) / "overwide-comparison-analysis.json"
+        overwide_path.write_text(
+            json.dumps(overwide_analysis, ensure_ascii=False), encoding="utf-8"
+        )
+        overwide_run = Path(temp) / "overwide-run"
+        run(command(
+            "prepare-comparison",
+            "--results", str(results),
+            "--run-dir", str(overwide_run),
+        ))
+        overwide_failure = run_failed(command(
+            "complete-comparison",
+            "--results", str(results),
+            "--analysis", str(overwide_path),
+            "--run-dir", str(overwide_run),
+        ))
+        assert "原子性门禁" in json.dumps(overwide_failure, ensure_ascii=False)
+        assert not (overwide_run / "comparison-gate.json").exists()
+        assert "打开各方答案聚合（未核验）" in comparison_stage["checkpoint"]["message"]
+        assert (run_dir / "02-comparison-report.html").exists()
+        prepared = run(command("prepare-authority", "--run-dir", str(run_dir)), keyless_environment)
+        assert prepared["taskCount"] == 1
+        assert prepared["status"] == "prepared" and prepared["trustedSearchRequiredCount"] == 0
+        request = json.loads((run_dir / "authority" / "requests" / "K1.json").read_text(encoding="utf-8"))
+        assert set(request["cloudPayload"]) == {"title", "knowledgePoint", "differingClaims"}
+        assert "answerMarkdown" not in json.dumps(request["cloudPayload"], ensure_ascii=False)
+        searched = run(command("search-authority", "--run-dir", str(run_dir), "--max-workers", "12"), keyless_environment)
+        assert searched["trustedSearchRequestCount"] == 0 and searched["dknowExemptCount"] == 1
+        assessments = run_dir / "authority" / "assessments"
+        assessments.mkdir(parents=True)
+        assessment = json.loads((AUTHORITY_FIXTURES / "K1-assessment.json").read_text(encoding="utf-8"))
+        (assessments / "K1.json").write_text(json.dumps(assessment, ensure_ascii=False), encoding="utf-8")
+        original_pipeline_run = fact_check_x.run
+
+        def fail_authority_report(arguments, environment=None):
+            if any(Path(str(item)).name == "render_authority_report.py" for item in arguments):
+                raise fact_check_x.PipelineError("强制报告渲染失败")
+            return original_pipeline_run(arguments, environment)
+
+        with patch.object(fact_check_x, "run", side_effect=fail_authority_report):
+            try:
+                fact_check_x.finalize_authority(
+                    type("Args", (), {"run_dir": str(run_dir), "assessments_dir": None})(),
+                    {key: Path(value) for key, value in located["skills"].items()},
+                )
+            except fact_check_x.PipelineError as exc:
+                assert "强制报告渲染失败" in str(exc)
+            else:
+                raise AssertionError("报告渲染失败应回滚权威核验事务")
+        assert json.loads(
+            (run_dir / "authority-gate.json").read_text(encoding="utf-8")
+        )["status"] == "searched"
+        assert not list((run_dir / "authority" / "results").glob("*.json"))
+        assert not (run_dir / "verification.json").exists()
+        assert not (run_dir / "03-authority-report.html").exists()
+        finalized = run(command("finalize-authority", "--run-dir", str(run_dir)))
+        assert finalized["status"] == "completed"
+        assert finalized["stage"] == "authority_completed"
+        assert finalized["deliverables"][0]["path"] == str(
+            (run_dir / "03-authority-report.html").resolve()
+        )
+        assert_deliverable(
+            finalized["deliverables"][0],
+            "权威核验后的最终答案",
+            run_dir / "03-authority-report.html",
+        )
+        assert finalized["checkpoint"]["mustPresentBeforeNextStage"] is True
+        assert finalized["checkpoint"]["path"] == str(
+            (run_dir / "03-authority-report.html").resolve()
+        )
+        assert "打开权威核验后的最终答案" in finalized["checkpoint"]["message"]
+        assert (run_dir / "03-authority-report.html").exists()
+        assert json.loads((run_dir / "authority-gate.json").read_text(encoding="utf-8"))["status"] == "finalized"
+        reopened = run(command(
+            "reopen-authority",
+            "--run-dir", str(run_dir),
+            "--reason", "修正被截断的权威结论",
+        ))
+        assert reopened["stage"] == "authority_reopened"
+        assert reopened["revision"] == 1
+        revision_dir = run_dir / "authority" / "revisions" / "revision-001"
+        assert (revision_dir / "revision.json").is_file()
+        assert (revision_dir / "verification.json").is_file()
+        assert (revision_dir / "authority" / "results" / "K1.json").is_file()
+        assert not list((run_dir / "authority" / "results").glob("*.json"))
+        assert not (run_dir / "verification.json").exists()
+        assert not (run_dir / "03-authority-report.html").exists()
+        reopened_gate = json.loads((run_dir / "authority-gate.json").read_text(encoding="utf-8"))
+        assert reopened_gate["status"] == "searched"
+        comparison_data = json.loads((run_dir / "comparison.json").read_text(encoding="utf-8"))
+        exact_title = comparison_data["knowledgePoints"][0]["description"]
+        assessment["authoritativeFinding"] = f"{exact_title}：每人每月最高提取1400元。"
+        (assessments / "K1.json").write_text(
+            json.dumps(assessment, ensure_ascii=False), encoding="utf-8"
+        )
+        refinalized = run(command("finalize-authority", "--run-dir", str(run_dir)))
+        assert refinalized["status"] == "completed"
+        reopened_verification = json.loads(
+            (run_dir / "verification.json").read_text(encoding="utf-8")
+        )
+        assert reopened_verification["finalAnswer"]["items"][0]["answer"] == assessment["authoritativeFinding"]
+        locked_verification = (run_dir / "verification.json").read_bytes()
+        locked_authority_report = (run_dir / "03-authority-report.html").read_bytes()
+        (run_dir / "verification.json").write_bytes(locked_verification + b"\n")
+        immutable_rejection = run_failed(command(
+            "deliver", "--results", str(results), "--run-dir", str(run_dir)
+        ))
+        assert "第三步权威核验结果已被修改" in immutable_rejection["error"]
+        (run_dir / "verification.json").write_bytes(locked_verification)
+        delivered = run(command("deliver", "--results", str(results), "--run-dir", str(run_dir)))
+        assert delivered["status"] == "completed"
+        assert delivered["trustedSearchRequestCount"] == 0 and delivered["dknowExemptCount"] == 1
+        assert delivered["answerReferenceReport"] == str((run_dir / "capture" / "report.html").resolve())
+        assert delivered["comparisonReport"] == str((run_dir / "comparison.html").resolve())
+        assert delivered["artifacts"]["report"] == str((run_dir / "report.html").resolve())
+        deliverable_paths = [item["path"] for item in delivered["deliverables"]]
+        assert deliverable_paths[:4] == [
+            str((run_dir / "01-capture-report.html").resolve()),
+            str((run_dir / "02-comparison-report.html").resolve()),
+            str((run_dir / "03-authority-report.html").resolve()),
+            str((run_dir / "04-final-report.html").resolve()),
+        ]
+        assert deliverable_paths[4] == delivered["artifacts"]["reportPackage"]
+        with zipfile.ZipFile(deliverable_paths[4]) as report_archive:
+            packaged_checkpoints = json.loads(
+                report_archive.read(
+                    "fact-check-x-report/data/stage-checkpoints.json"
+                ).decode("utf-8")
+            )
+        assert list(packaged_checkpoints["stages"]) == [
+            "capture", "comparison", "authority", "evaluation"
+        ]
+        assert packaged_checkpoints["stages"]["evaluation"]["status"] == "completed"
+        for packaged_stage in packaged_checkpoints["stages"].values():
+            for packaged_item in packaged_stage.get("deliverables") or []:
+                assert packaged_item["fileUri"].startswith("../")
+                assert "file://" not in packaged_item["markdownLink"]
+                assert "file://" not in packaged_item["message"]
+        assert delivered["checkpoint"]["mustPresentBeforeNextStage"] is True
+        assert delivered["checkpoint"]["path"] == str(
+            (run_dir / "04-final-report.html").resolve()
+        )
+        assert "打开各方答案测评报告" in delivered["checkpoint"]["message"]
+        assert (run_dir / "verification.json").read_bytes() == locked_verification
+        assert (run_dir / "03-authority-report.html").read_bytes() == locked_authority_report
+        assert all(Path(item["path"]).exists() for item in delivered["deliverables"])
+        for item, label, path in (
+            (delivered["deliverables"][0], "各方答案汇总", run_dir / "01-capture-report.html"),
+            (delivered["deliverables"][1], "各方答案聚合（未核验）", run_dir / "02-comparison-report.html"),
+            (delivered["deliverables"][2], "权威核验后的最终答案", run_dir / "03-authority-report.html"),
+            (delivered["deliverables"][3], "各方答案测评报告", run_dir / "04-final-report.html"),
+            (delivered["deliverables"][4], "完整可分发报告包", Path(deliverable_paths[4])),
+        ):
+            assert_deliverable(item, label, path)
+        manifest = json.loads((run_dir / "pipeline.json").read_text(encoding="utf-8"))
+        verification = json.loads((run_dir / "verification.json").read_text(encoding="utf-8"))
+        report = (run_dir / "report.html").read_text(encoding="utf-8")
+        authority_report = (run_dir / "03-authority-report.html").read_text(
+            encoding="utf-8"
+        )
+        comparison_deliverable = (run_dir / "02-comparison-report.html").read_text(
+            encoding="utf-8"
+        )
+        assert manifest["schemaVersion"] == "fact-check-x/pipeline@2"
+        assert verification["finalAnswer"]["status"] == "verified"
+        assert verification["finalAnswer"]["knowledgePointIds"] == ["K1"]
+        assert verification["knowledgePoints"][0]["authority"]["verdicts"]["doubao"]["category"] == "misleading"
+        assert "权威核验报告（问题：" in authority_report
+        assert "权威核验后的最终答案" in authority_report
+        assert "各平台裁决" not in authority_report
+        assert "data-fcx-authority-binding-sha256" not in authority_report
+        report_names = [
+            "01-capture-report.html",
+            "02-comparison-report.html",
+            "03-authority-report.html",
+            "04-final-report.html",
+        ]
+        for report_name in report_names:
+            report_html = (run_dir / report_name).read_text(encoding="utf-8")
+            assert report_html.count('data-fcx-report-nav="1"') == 1
+            assert 'aria-current="page"' in report_html
+            assert "来源链接待补" not in report_html
+            assert all(
+                f'href="{target}"' in report_html
+                for target in report_names
+                if target != report_name
+            )
+        assert all(
+            f'href="{target}"' in comparison_deliverable
+            for target in (
+                "01-capture-report.html",
+                "03-authority-report.html",
+                "04-final-report.html",
+            )
+        )
+        assert "① 平台表现概览" in report
+        assert "② 逐知识点核验明细" in report
+        assert "③ 原始答案与参考文献（存证）" in report
+        assert "④ 指标口径速查" in report
+        assert "⑤ 评测元信息" in report
+        assert report.count("② 逐知识点核验明细") == 1
+        assert 'data-fcx-locked-role="direct"' in report
+        assert 'data-fcx-locked-role="reference"' in report
+        assert "② 直接答案逐知识点测评" not in report
+        assert "③ 补充参考逐知识点测评" not in report
+        assert "② 直接答案逐条判定" not in report
+        assert "②-补 补充参考分析" not in report
+        assert (run_dir / "comparison.html").exists()
+        capture_report = (run_dir / "capture" / "report.html").read_text(encoding="utf-8")
+        assert "各方答案汇总（问题：" in capture_report
+        assert "每人每月最高提取 1400 元" in capture_report
+        assert (run_dir / "capture" / "report.md").exists()
+        assert manifest["artifacts"]["answerReferenceReport"] == str((run_dir / "capture" / "report.html").resolve())
+        assert manifest["artifacts"]["captureGate"] == str((run_dir / "capture-gate.json").resolve())
+        assert manifest["artifacts"]["comparisonAnalysis"] == str(
+            (run_dir / "comparison-analysis.json").resolve()
+        )
+        assert manifest["artifacts"]["authorityGate"] == str((run_dir / "authority-gate.json").resolve())
+
+        artifact_source = Path(temp) / "artifact-source"
+        artifact_source.mkdir()
+        artifact_results = json.loads(results.read_text(encoding="utf-8"))
+        for platform in artifact_results["platforms"]:
+            platform_id = platform["platform"]
+            platform["artifacts"] = {
+                "screenshot": f"artifacts/{platform_id}/screenshot.png",
+                "html": f"artifacts/{platform_id}/page.html",
+            }
+            artifact_dir = artifact_source / "artifacts" / platform_id
+            artifact_dir.mkdir(parents=True)
+            (artifact_dir / "screenshot.png").write_bytes(f"png:{platform_id}".encode())
+            (artifact_dir / "page.html").write_text(
+                f"<html>{platform_id}</html>", encoding="utf-8"
+            )
+        artifact_results_path = artifact_source / "results.json"
+        artifact_results_path.write_text(
+            json.dumps(artifact_results, ensure_ascii=False), encoding="utf-8"
+        )
+        (artifact_source / "capture-recovery.json").write_text(
+            json.dumps({
+                "schemaVersion": "fact-check-x/capture-recovery@2",
+                "status": "completed",
+            }, ensure_ascii=False),
+            encoding="utf-8",
+        )
+        artifact_run = Path(temp) / "artifact-run"
+        run(command(
+            "prepare-comparison",
+            "--results",
+            str(artifact_results_path),
+            "--run-dir",
+            str(artifact_run),
+        ))
+        artifact_gate = json.loads(
+            (artifact_run / "capture-gate.json").read_text(encoding="utf-8")
+        )
+        assert len(artifact_gate["artifactSha256"]) == 4
+        assert artifact_gate["recoverySha256"]
+        assert (artifact_run / "capture/capture-recovery.json").is_file()
+        for platform in artifact_results["platforms"]:
+            platform_id = platform["platform"]
+            assert (
+                artifact_run / f"capture/artifacts/{platform_id}/screenshot.png"
+            ).is_file()
+            assert (
+                artifact_run / f"capture/artifacts/{platform_id}/page.html"
+            ).is_file()
+            assert (
+                artifact_run / f"artifacts/{platform_id}/screenshot.png"
+            ).is_file()
+            assert (
+                artifact_run / f"artifacts/{platform_id}/page.html"
+            ).is_file()
+        missing_artifact_results = json.loads(
+            artifact_results_path.read_text(encoding="utf-8")
+        )
+        (
+            artifact_source
+            / missing_artifact_results["platforms"][0]["artifacts"]["screenshot"]
+        ).unlink()
+        missing_artifact_run = Path(temp) / "missing-artifact-run"
+        missing_artifact = run_failed(command(
+            "prepare-comparison",
+            "--results",
+            str(artifact_results_path),
+            "--run-dir",
+            str(missing_artifact_run),
+        ))
+        assert "存证文件不存在" in missing_artifact["error"]
+
+        configuration_run = Path(temp) / "configuration-run"
+        run(command("prepare-comparison", "--results", str(results), "--run-dir", str(configuration_run)))
+        configuration_analysis = json.loads((COMPARE_FIXTURES / "comparison-analysis.json").read_text(encoding="utf-8"))
+        configuration_analysis["knowledgePoints"][0]["trustedAnchor"] = {"eligible": False}
+        configuration_analysis["knowledgePoints"][0]["claims"]["dknowc-chat"] = {
+            "covered": False,
+            "claim": "",
+            "answerExcerpt": "",
+            "citedReferenceIndexes": [],
+            "answerLevelReferenceIndexes": [],
+            "faithfulness": "insufficient",
+            "reason": "",
+            "evidence": [],
+        }
+        configuration_analysis_path = configuration_run / "comparison-analysis.json"
+        configuration_analysis_path.write_text(json.dumps(configuration_analysis, ensure_ascii=False), encoding="utf-8")
+        run(command("complete-comparison", "--results", str(results), "--analysis", str(configuration_analysis_path), "--run-dir", str(configuration_run)))
+        configuration_required = run_failed(command(
+            "prepare-authority",
+            "--run-dir",
+            str(configuration_run),
+        ), keyless_environment)
+        assert configuration_required["status"] == "configuration_required"
+        assert configuration_required["action"] == "configure_trusted_search"
+        assert configuration_required["trustedSearchRequiredCount"] == 1
+        assert "您只需完成登录" in configuration_required["userPrompt"]
+        assert "自动读取已有的可用 Key" in configuration_required["userPrompt"]
+        assert configuration_required["configuration"]["providerUrl"] == "https://platform.dknowc.cn/auth/#/login"
+        assert configuration_required["configuration"]["interaction"] == "browser_login_only"
+        assert configuration_required["configuration"]["requiresChatSecret"] is False
+        assert configuration_required["configuration"]["autoResume"] is True
+        assert configuration_required["configuration"]["sharedAcrossCarriers"] is True
+        assert configuration_required["configuration"]["command"][-1] == "configure"
+        assert configuration_required["configuration"]["command"][-2].endswith(
+            "trusted_search_config.py"
+        )
+        assert json.loads((configuration_run / "authority-gate.json").read_text(encoding="utf-8"))["status"] == "configuration_required"
+        search_blocked = run_failed(command(
+            "search-authority",
+            "--run-dir",
+            str(configuration_run),
+            "--max-workers",
+            "12",
+        ), keyless_environment)
+        assert search_blocked["status"] == "configuration_required"
+        assert search_blocked["action"] == "configure_trusted_search"
+        assert not (configuration_run / "authority" / "evidence" / "batch.json").exists()
+        bypass_attempt = run_failed(command(
+            "finalize-authority",
+            "--run-dir",
+            str(configuration_run),
+        ), keyless_environment)
+        assert bypass_attempt["status"] == "failed"
+        assert "禁止写入裁决" in bypass_attempt["error"]
+        bypass_delivery = run_failed(command(
+            "deliver",
+            "--results",
+            str(results),
+            "--run-dir",
+            str(configuration_run),
+        ), keyless_environment)
+        assert bypass_delivery["status"] == "failed"
+        assert "缺少 authority 阶段产物交付记录" in bypass_delivery["error"]
+
+        tamper_run = Path(temp) / "tamper-run"
+        run(command("prepare-comparison", "--results", str(results), "--run-dir", str(tamper_run)))
+        run(command("complete-comparison", "--results", str(results), "--analysis", str(COMPARE_FIXTURES / "comparison-analysis.json"), "--run-dir", str(tamper_run)))
+        tampered = json.loads((tamper_run / "comparison.json").read_text(encoding="utf-8"))
+        tampered["knowledgePoints"][0]["claims"]["doubao"]["sourceLevel"] = "official"
+        (tamper_run / "comparison.json").write_text(json.dumps(tampered, ensure_ascii=False), encoding="utf-8")
+        tamper_rejected = run_failed(command("prepare-authority", "--run-dir", str(tamper_run)))
+        assert "被修改" in tamper_rejected["error"]
+
+        pollution_run = Path(temp) / "pollution-run"
+        run(command("prepare-comparison", "--results", str(results), "--run-dir", str(pollution_run)))
+        run(command("complete-comparison", "--results", str(results), "--analysis", str(COMPARE_FIXTURES / "comparison-analysis.json"), "--run-dir", str(pollution_run)))
+        run(command("prepare-authority", "--run-dir", str(pollution_run)))
+        run(command("search-authority", "--run-dir", str(pollution_run), "--max-workers", "12"))
+        pollution_assessments = pollution_run / "authority" / "assessments"
+        pollution_assessments.mkdir(parents=True)
+        shutil.copyfile(AUTHORITY_FIXTURES / "K1-assessment.json", pollution_assessments / "K1.json")
+        pollution_results = pollution_run / "authority" / "results"
+        pollution_results.mkdir(parents=True)
+        (pollution_results / "KP-001.json").write_text("{}\n", encoding="utf-8")
+        pollution_rejected = run_failed(command("finalize-authority", "--run-dir", str(pollution_run)))
+        assert "results 目录必须为空" in pollution_rejected["error"]
+
+        invalid_assessments = Path(temp) / "invalid-assessments"
+        invalid_assessments.mkdir()
+        (invalid_assessments / "K1.json").write_text(json.dumps({
+            "requestId": "K1",
+            "verdict": "supported",
+            "officialAnswer": "旧结构结论",
+            "platformAssessment": {
+                "dknowc-chat": {"verdict": "supported", "reason": "旧结构"},
+                "doubao": {"verdict": "supported", "reason": "旧结构"},
+            },
+        }, ensure_ascii=False), encoding="utf-8")
+        invalid_run = Path(temp) / "invalid-run"
+        run(command("prepare-comparison", "--results", str(results), "--run-dir", str(invalid_run)))
+        run(command("complete-comparison", "--results", str(results), "--analysis", str(COMPARE_FIXTURES / "comparison-analysis.json"), "--run-dir", str(invalid_run)))
+        run(command("prepare-authority", "--run-dir", str(invalid_run)))
+        run(command("search-authority", "--run-dir", str(invalid_run), "--max-workers", "12"))
+        invalid_final = run_failed(command(
+            "finalize-authority",
+            "--run-dir",
+            str(invalid_run),
+            "--assessments-dir",
+            str(invalid_assessments),
+        ))
+        assert invalid_final["status"] == "failed"
+        assert "裁决文件结构错误" in invalid_final["error"]
+        assert not (invalid_run / "authority" / "results" / "K1.json").exists()
+
+        review_run = Path(temp) / "review-run"
+        run(command("prepare-comparison", "--results", str(results), "--run-dir", str(review_run)))
+        run(command("complete-comparison", "--results", str(results), "--analysis", str(COMPARE_FIXTURES / "comparison-analysis.json"), "--run-dir", str(review_run)))
+        run(command("prepare-authority", "--run-dir", str(review_run)))
+        run(command("search-authority", "--run-dir", str(review_run), "--max-workers", "12"))
+        review_assessments = review_run / "authority" / "assessments"
+        review_assessments.mkdir(parents=True)
+        review_assessment = json.loads((AUTHORITY_FIXTURES / "K1-assessment.json").read_text(encoding="utf-8"))
+        review_assessment["verdicts"]["doubao"] = {
+            "verdict": "insufficient",
+            "reason": "当前证据不足以完成裁决",
+            "evidenceIds": [],
+        }
+        (review_assessments / "K1.json").write_text(json.dumps(review_assessment, ensure_ascii=False), encoding="utf-8")
+        review_final = run(command("finalize-authority", "--run-dir", str(review_run)))
+        assert review_final["status"] == "completed"
+        assert review_final["stage"] == "authority_completed"
+        assert review_final["evidenceGapCount"] == 1
+        assert review_final["deliverables"][0]["path"] == str(
+            (review_run / "03-authority-report.html").resolve()
+        )
+        review_authority_report = (
+            review_run / "03-authority-report.html"
+        ).read_text(encoding="utf-8")
+        assert "证据不足项" in review_authority_report
+        assert 'href="04-final-report.html"' in review_authority_report
+        assert json.loads((review_run / "authority-gate.json").read_text(encoding="utf-8"))["status"] == "finalized"
+        review_delivery = run(command("deliver", "--results", str(results), "--run-dir", str(review_run)))
+        assert review_delivery["status"] == "completed"
+        assert review_delivery["evidenceGapCount"] == 1
+        assert (review_run / "04-final-report.html").is_file()
+        review_verification = json.loads(
+            (review_run / "verification.json").read_text(encoding="utf-8")
+        )
+        assert review_verification["status"] == "completed"
+        assert review_verification["evidenceGapCount"] == 1
+        assert review_verification["evidenceGaps"][0]["platform"] == "doubao"
+
+        failed_results = json.loads(results.read_text(encoding="utf-8"))
+        failed_results["platforms"][0]["answerMarkdown"] = "为您智能匹配到当前所在区域为“北京市”，如想咨询其他区域可点击修改"
+        failed_results["platforms"][1]["status"] = "failed"
+        failed_results["platforms"][1]["answerMarkdown"] = ""
+        failed_results["platforms"][1]["error"] = "No answer text detected."
+        failed_path = Path(temp) / "failed-results.json"
+        failed_path.write_text(json.dumps(failed_results, ensure_ascii=False), encoding="utf-8")
+        failed_run = Path(temp) / "failed-run"
+        rejected = run_failed(command("prepare-comparison", "--results", str(failed_path), "--run-dir", str(failed_run)))
+        assert rejected["status"] == "failed"
+        assert "禁止进入知识点对比" in rejected["error"]
+        assert not (failed_run / "comparison-task.json").exists()
+
+        answered_with_region_footer = json.loads(results.read_text(encoding="utf-8"))
+        answered_with_region_footer["platforms"][0]["answerMarkdown"] = (
+            "北京市高考报名资格、网上申请、填报缴费和现场确认的完整政策回答。" * 10
+            + "页面底部：为您智能匹配到当前所在区域为“北京市”，如想咨询其他区域可点击修改"
+        )
+        answered_with_region_footer_path = Path(temp) / "answered-with-region-footer.json"
+        answered_with_region_footer_path.write_text(
+            json.dumps(answered_with_region_footer, ensure_ascii=False),
+            encoding="utf-8",
+        )
+        answered_with_region_footer_run = Path(temp) / "answered-with-region-footer-run"
+        accepted_with_footer = run(command(
+            "prepare-comparison",
+            "--results",
+            str(answered_with_region_footer_path),
+            "--run-dir",
+            str(answered_with_region_footer_run),
+        ))
+        assert accepted_with_footer["stage"] == "capture_completed"
+        assert (answered_with_region_footer_run / "comparison-task.json").exists()
+
+        recovery_capture = Path(temp) / "recovery-capture"
+        recovery_capture.mkdir()
+        recovery_results = recovery_capture / "results.json"
+        recovery_results.write_text(results.read_text(encoding="utf-8"), encoding="utf-8")
+        (recovery_capture / "capture-recovery.json").write_text(
+            json.dumps({
+                "schemaVersion": "fact-check-x/capture-recovery@2",
+                "status": "required",
+                "action": "computer_use",
+            }, ensure_ascii=False),
+            encoding="utf-8",
+        )
+        recovery_run = Path(temp) / "recovery-run"
+        recovery_rejected = run_failed(command(
+            "prepare-comparison",
+            "--results",
+            str(recovery_results),
+            "--run-dir",
+            str(recovery_run),
+        ))
+        assert "Computer Use" in recovery_rejected["error"]
+        assert not (recovery_run / "comparison-task.json").exists()
+        assert not (ROOT / "assets").exists()
+    if os.getenv("FACT_CHECK_X_ASSERTIONS_OUTPUT"):
+        Path(os.environ["FACT_CHECK_X_ASSERTIONS_OUTPUT"]).write_text(json.dumps({
+            "schemaVersion": "fact-check-x/test-assertions@1",
+            "actualAssertionIds": [
+                "report.stage_artifacts_visible",
+                "report.checkpoints_indexed",
+                "report.unverified_draft_visible",
+                "report.verified_final_answer_visible",
+                "report.stage2_role_sections",
+                "report.stage4_single_locked_detail",
+                "report.technical_failure_notice",
+                "report.renamed_four_stages",
+                "authority.finalize_transaction_rollback",
+                "comparison.overwide_claim_blocks_authority",
+                "path.windows_drive_normalized",
+            ],
+        }), encoding="utf-8")
+    print("PASS Fact-Check-X 统一入口")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())