File size: 86,315 Bytes
c8d30bc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 | """
main.py - ThreatHunter 銝餌撘
==============================
Pipeline 嗆嚗v3.1嚗嚗Orchestrator [Layer 1 銝西] Scout Analyst Debate Advisor
嗆嚗v3.1 Orchestrator 撽嚗嚗
Orchestrator 頝舐梧
頝臬 A嚗憟隞嗆 頝喲 Security Guard
頝臬 B嚗摰渡撘蝣 Layer 1 銝西嚗Security Guard + Intel Fusion嚗
頝臬 C嚗隞嗅摹蝵 頝喲 Analyst + Debate
頝臬 D嚗擖鋆 芷頝雿靽∪ CVE
嗆嚗靽靘嚗嚗
雿輻刻頛詨 "Django 4.2, Redis 7.0"
main.py Pipeline
Scout 嗯 Analyst 嗯 Critic
(鈭撖行園) (函斗) (舀)
潑
Advisor
(蝯鋆瘙箏勗)
瘥 Stage 函嚗
- try-except + Graceful Degradation
- StepLogger 摮甇仿亥
- Harness 靽撅歹 agents/*.py 折剁
鞈瘚嚗
Scout 頛詨 (dict) Analyst 頛詨 (dict)
Analyst 頛詨 (dict) Critic 頛詨 (dict)
Analyst + Critic 頛詨 Advisor 頛詨 (dict)
Advisor 頛詨 (dict) + pipeline_meta 蝯蝯
Harness Engineering 靽嚗
- 瘥 Agent 賭蝙函撖行芋蝯嚗agents/*.py嚗嚗 Stub
- Critic ENABLE_CRITIC 批塚舀嚗
- 瘥 Stage 函 Graceful Degradation 蝝頝臬
- 函閮 Observability 亥嚗FINAL_PLAN.md 舀 2嚗
- 17 撅 Harness 靽撅歹 agents/*.py 折剁
萄嚗project_CONSTITUTION.md + HARNESS_ENGINEERING.md
"""
import json
import logging
import sys
import time
from concurrent.futures import ThreadPoolExecutor, as_completed
from datetime import datetime, timezone
from pathlib import Path
from typing import Any
from config import (
ENABLE_CRITIC,
degradation_status,
rate_limiter,
)
logger = logging.getLogger("threathunter.main")
import os # noqa: E402 (import after logger for ordering clarity)
# (comment encoding corrupted)
# (comment encoding corrupted)
# (comment encoding corrupted)
SANDBOX_ENABLED = os.getenv("SANDBOX_ENABLED", "true").lower() == "true"
try:
from sandbox.docker_sandbox import run_in_sandbox, is_docker_available
_DOCKER_SANDBOX_OK = True
except ImportError:
_DOCKER_SANDBOX_OK = False
def run_in_sandbox(*args, **kwargs): # type: ignore[misc]
return {"error": "SANDBOX_NOT_AVAILABLE", "fallback": True}
def is_docker_available() -> bool: # type: ignore[misc]
return False
if SANDBOX_ENABLED:
logger.info(
"[SANDBOX] Docker isolation ENABLED | docker_available=%s",
is_docker_available(),
)
else:
logger.debug("[SANDBOX] Docker isolation DISABLED (in-process mode)")
# (comment encoding corrupted)
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
class StepLogger:
"""瘥 Agent Stage 摮甇仿餈質馱具"""
def __init__(self, agent_name: str):
self.agent_name = agent_name
self.steps: list[dict[str, Any]] = []
def log(
self, step: str, status: str, detail: str = "", duration_ms: int = 0
) -> None:
entry = {
"step": step,
"agent": self.agent_name,
"timestamp": datetime.now(timezone.utc).isoformat(),
"status": status,
"detail": detail,
"duration_ms": duration_ms,
}
self.steps.append(entry)
icon = (
"OK" if status == "SUCCESS" else "WAIT" if status == "RUNNING" else "FAIL"
)
logger.info("[%s] %s | %s | %s", icon, self.agent_name.upper(), step, detail)
def summary(self) -> dict[str, Any]:
failed = [s for s in self.steps if s["status"] == "FAILED"]
return {
"agent": self.agent_name,
"total_steps": len(self.steps),
"failed_steps": len(failed),
"steps": self.steps,
}
def _normalize_pipeline_task_plan(task_plan: dict[str, Any]) -> dict[str, Any]:
"""
Canonical runtime DAG:
Security Guard + Scout discover in parallel, then Intel Fusion ranks risk.
"""
plan = dict(task_plan or {})
scan_path = plan.get("path", "B")
if scan_path == "B":
plan["parallel_layer1"] = ["security_guard", "scout"]
plan["fusion_after_discovery"] = True
plan["agents_to_run"] = [
"security_guard", "scout", "intel_fusion", "analyst", "debate", "judge"
]
elif scan_path == "A":
plan["parallel_layer1"] = ["scout"]
plan["fusion_after_discovery"] = True
plan["agents_to_run"] = ["scout", "intel_fusion", "analyst", "debate", "judge"]
return plan
_L0_TO_RUNTIME_INPUT_TYPE = {
"package_list": "pkg",
"source_code": "code",
"mixed": "code",
"config_file": "config",
"sql_review": "sql_review",
"blocked": "injection",
}
def _canonical_runtime_input_type(requested: str, l0_input_type: str = "") -> str:
"""用 L0 deterministic 分類修正 UI/API 預設值,避免 code 被當 package。"""
runtime = (requested or "pkg").strip().lower()
if runtime not in {"pkg", "code", "config", "injection", "sql_review"}:
runtime = "pkg"
l0_runtime = _L0_TO_RUNTIME_INPUT_TYPE.get(str(l0_input_type or "").strip().lower(), "")
if runtime == "pkg" and l0_runtime in {"code", "config", "injection", "sql_review"}:
return l0_runtime
return runtime
def _constrain_task_plan_by_input_type(task_plan: dict[str, Any], input_type: str) -> dict[str, Any]:
"""L0 類型是硬約束;Orchestrator 不可把 source code 誤路由成 Path C/A。"""
plan = dict(task_plan or {})
original_path = plan.get("path", "B")
if input_type in {"code", "injection"} and original_path != "B":
plan["path"] = "B"
plan["_route_corrected"] = {
"from": original_path,
"to": "B",
"reason": f"l0_input_type={input_type}",
}
elif input_type == "sql_review":
if original_path != "C":
plan["_route_corrected"] = {
"from": original_path,
"to": "C",
"reason": "l0_input_type=sql_review",
}
plan["path"] = "C"
plan["parallel_layer1"] = []
plan["fusion_after_discovery"] = False
plan["agents_to_run"] = ["sql_review"]
elif input_type == "config" and original_path != "C":
plan["path"] = "C"
plan["_route_corrected"] = {
"from": original_path,
"to": "C",
"reason": "l0_input_type=config",
}
return _normalize_pipeline_task_plan(plan)
def _load_dim11_expected_config() -> dict[str, Any]:
"""載入 DIM11 benchmark gate;一般掃描沒有此檔或沒有 exact match 時不啟用。"""
expected_path = Path(__file__).resolve().parent / "tests" / "fixtures" / "dim11_redteam" / "expected_results.json"
try:
return json.loads(expected_path.read_text(encoding="utf-8"))
except (OSError, json.JSONDecodeError, ValueError) as exc:
logger.debug("[DIM11] benchmark config unavailable: %s", exc)
return {}
def _build_dim11_benchmark_context(
source: str,
input_type: str,
task_plan: dict[str, Any],
code_patterns: list[dict[str, Any]],
scout_output: dict[str, Any],
extracted_packages: list[str],
) -> dict[str, Any] | None:
"""只對 DIM11 fixture exact match 啟用 expected-CWE recall challenge。"""
if input_type != "code":
return None
config = _load_dim11_expected_config()
fixture_dir = Path(__file__).resolve().parent / "tests" / "fixtures" / "dim11_redteam"
observed_cwes = sorted({
str(pattern.get("cwe_id", ""))
for pattern in code_patterns
if str(pattern.get("cwe_id", "")).startswith("CWE-")
})
for case in config.get("fixtures", []):
expected_cwes = case.get("expected_cwe_categories")
if not expected_cwes or case.get("exclude_from_cwe_recall"):
continue
fixture_name = str(case.get("fixture", ""))
try:
fixture_source = (fixture_dir / fixture_name).read_text(encoding="utf-8", errors="replace")
except OSError:
continue
if source != fixture_source:
continue
external_findings = scout_output.get("vulnerabilities", [])
external_pollution_count = 0
if case.get("external_requires_package_evidence") and external_findings and not extracted_packages:
external_pollution_count = len(external_findings)
expected_path = str(case.get("expected_path", ""))
actual_path = str(task_plan.get("path", ""))
return {
"benchmark": "dim11_redteam",
"fixture": fixture_name,
"expected_cwe_categories": sorted(set(expected_cwes)),
"observed_cwe_categories": observed_cwes,
"min_category_recall": float(case.get("min_category_recall", 0.7)),
"route_correct": actual_path == expected_path,
"expected_path": expected_path,
"actual_path": actual_path,
"external_pollution_count": external_pollution_count,
"activation": "exact_fixture_match",
}
return None
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
def stage_scout(
tech_stack: str,
input_type: str = "pkg",
intel_fusion_result: dict[str, Any] | None = None,
) -> tuple[dict[str, Any], StepLogger]:
"""
Stage 1: Scout Agent 萄瞍瘣
雿輻 agents/scout.py 撖血祕雿
Graceful Degradation: 憭望喟征鞈嚗霈敺蝥 Agent 仿舐洵銝甈~
v3.7: input_type 瘙箏頛芸 Skill SOP (Path-Aware Skills)
Returns:
(result_dict, step_logger) 頛詨箏亥餈質馱
"""
from agents.scout import run_scout_pipeline
sl = StepLogger("scout")
sl.log("INIT", "RUNNING", f"tech_stack={tech_stack} | input_type={input_type}")
t0 = time.time()
try:
result = run_scout_pipeline(
tech_stack,
input_type=input_type,
intel_fusion_result=intel_fusion_result,
)
duration_ms = int((time.time() - t0) * 1000)
vuln_count = len(result.get("vulnerabilities", []))
sl.log(
"COMPLETE", "SUCCESS", f"found {vuln_count} vulnerabilities", duration_ms
)
return result, sl
except Exception as e:
duration_ms = int((time.time() - t0) * 1000)
# (comment encoding corrupted)
is_rate_limit = "429" in str(e) or "rate limit" in str(e).lower()
if is_rate_limit:
logger.warning("[SCOUT] Rate limited returning empty results (not a real failure)")
sl.log("COMPLETE", "RATE_LIMITED", str(e)[:100], duration_ms)
else:
logger.error("Scout Stage failed: %s", e)
sl.log("COMPLETE", "FAILED", str(e)[:100], duration_ms)
degradation_status.degrade("Scout", str(e))
# (comment encoding corrupted)
return {
"scan_id": f"scan_degraded_{int(time.time())}",
"vulnerabilities": [],
"summary": {"total": 0, "critical": 0, "high": 0, "medium": 0, "low": 0},
"_degraded": not is_rate_limit, # 429 銝蝞甇蝝
"_error": str(e),
}, sl
def stage_intel_fusion(
discovery_input: dict[str, Any] | list[str] | str,
on_progress: Any = None,
) -> tuple[dict[str, Any], StepLogger]:
"""
Stage: Intel Fusion runs after Scout and Security Guard discovery.
It receives normalized findings instead of raw source code.
"""
from agents.intel_fusion import run_intel_fusion
sl = StepLogger("intel_fusion")
sl.log("INIT", "RUNNING", "post-discovery fusion")
t0 = time.time()
try:
result = run_intel_fusion(discovery_input, on_progress)
duration_ms = int((time.time() - t0) * 1000)
result.setdefault("_duration_ms", duration_ms)
fusion_count = len(result.get("fusion_results", []))
sl.log("COMPLETE", "SUCCESS", f"scored {fusion_count} findings", duration_ms)
return result, sl
except Exception as e:
duration_ms = int((time.time() - t0) * 1000)
logger.error("Intel Fusion Stage failed: %s", e)
sl.log("COMPLETE", "FAILED", str(e)[:100], duration_ms)
degradation_status.degrade("Intel Fusion", str(e))
return {
"fusion_results": [],
"strategy_applied": "degraded",
"api_health_summary": {},
"_degraded": True,
"_error": str(e),
"_duration_ms": duration_ms,
}, sl
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
def stage_analyst(
scout_output: dict[str, Any],
input_type: str = "pkg",
) -> tuple[dict[str, Any], StepLogger]:
"""
Stage 2: Analyst Agent vulnerability chain analysis.
v3.7: input_type selects path-aware chain analysis skill.
Returns:
(result_dict, step_logger)
"""
from agents.analyst import run_analyst_pipeline
sl = StepLogger("analyst")
sl.log(
"INIT", "RUNNING",
f"input_vulns={len(scout_output.get('vulnerabilities', []))} | input_type={input_type}"
)
t0 = time.time()
try:
result = run_analyst_pipeline(scout_output, input_type=input_type)
duration_ms = int((time.time() - t0) * 1000)
risk = result.get("risk_score", 0)
sl.log("COMPLETE", "SUCCESS", f"risk_score={risk}", duration_ms)
return result, sl
except Exception as e:
duration_ms = int((time.time() - t0) * 1000)
logger.error("Analyst Stage failed: %s", e)
sl.log("COMPLETE", "FAILED", str(e)[:100], duration_ms)
degradation_status.degrade("Analyst", str(e))
# (comment encoding corrupted)
vulns = scout_output.get("vulnerabilities", [])
analysis = [
{
"cve_id": v.get("cve_id", "UNKNOWN"),
"original_cvss": v.get("cvss_score", 0),
"adjusted_risk": v.get("severity", "UNKNOWN"),
"in_cisa_kev": False,
"exploit_available": False,
"chain_risk": {
"is_chain": False,
"chain_with": [],
"chain_description": "chain_analysis: SKIPPED",
"confidence": "NEEDS_VERIFICATION",
},
"reasoning": "Analyst degraded - chain analysis skipped.",
}
for v in vulns
]
# (comment encoding corrupted)
code_patterns = scout_output.get("code_patterns", [])
for cp in code_patterns:
sev = cp.get("severity", "MEDIUM")
cvss_equiv = 9.0 if sev == "CRITICAL" else (7.5 if sev == "HIGH" else 5.0)
analysis.append({
"finding_id": cp.get("finding_id", "CODE-000"),
"cve_id": None,
"pattern_type": cp.get("pattern_type", "UNKNOWN"),
"cwe_id": cp.get("cwe_id", "CWE-unknown"),
"owasp_category": cp.get("owasp_category", ""),
"severity": sev,
"snippet": cp.get("snippet", ""),
"line_no": cp.get("line_no", 0),
"original_cvss": cvss_equiv,
"adjusted_risk": sev,
"in_cisa_kev": False,
"exploit_available": False,
"chain_risk": {
"is_chain": False,
"chain_with": [],
"chain_description": "Analyst degraded - deterministic pattern only.",
"confidence": "HIGH", # 蝣箏批菜葫嚗靽∪擃
},
"reasoning": f"Deterministic detection: {cp.get('pattern_type')} ({cp.get('cwe_id')}). Analyst degraded but code pattern is confirmed by Security Guard.",
})
if code_patterns:
logger.info("[DEGRADED] Analyst fallback preserved %d code_patterns", len(code_patterns))
return {
"scan_id": scout_output.get("scan_id", "unknown"),
"risk_score": 50,
"risk_trend": "+0",
"analysis": analysis,
"_degraded": True,
"_error": str(e),
}, sl
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
def stage_critic(
analyst_output: dict[str, Any],
input_type: str = "pkg",
) -> tuple[dict[str, Any], StepLogger]:
"""
Stage 3: Adversarial Debate Engine.
v5.3: 升級為 3 輪辯論引擎(Du et al. 2023 ICML, arXiv:2305.14325)。
無共識時由 Judge sub-agent 仲裁。偏保守原則:高估風險比低估安全。
Returns:
(result_dict, step_logger)
"""
from agents.debate_engine import run_debate_pipeline
sl = StepLogger("critic")
sl.log("INIT", "RUNNING", f"enable_critic={ENABLE_CRITIC} | input_type={input_type} | mode=3-round-debate")
t0 = time.time()
try:
result = run_debate_pipeline(
analyst_output,
input_type=input_type,
on_progress=None,
)
duration_ms = int((time.time() - t0) * 1000)
verdict = result.get("verdict", "UNKNOWN")
score = result.get("weighted_score", 0)
meta = result.get("_debate_meta", {})
consensus = meta.get("consensus", None)
rounds = meta.get("total_rounds", "?")
judge_invoked = meta.get("judge_invoked", False)
sl.log(
"COMPLETE", "SUCCESS",
f"verdict={verdict} score={score} consensus={consensus} rounds={rounds} judge={judge_invoked}",
duration_ms,
)
return result, sl
except Exception as e:
duration_ms = int((time.time() - t0) * 1000)
logger.error("Debate Stage failed: %s", e)
sl.log("COMPLETE", "FAILED", str(e)[:100], duration_ms)
degradation_status.degrade("Critic", str(e))
return {
"debate_rounds": 0,
"challenges": [],
"scorecard": {
"evidence": 0.6,
"chain_completeness": 0.6,
"critique_quality": 0.6,
"defense_quality": 0.6,
"calibration": 0.6,
},
"weighted_score": 60.0,
"verdict": "MAINTAIN",
"reasoning": "Debate engine degraded - all rounds skipped.",
"_degraded": True,
"_error": str(e),
}, sl
def stage_advisor(
analyst_output: dict[str, Any],
critic_output: dict[str, Any],
input_type: str = "pkg",
) -> tuple[dict[str, Any], StepLogger]:
"""
Stage 4: Advisor Agent action report generation.
v3.7: input_type selects path-aware action report skill.
Returns:
(result_dict, step_logger)
"""
from agents.advisor import run_advisor_pipeline
sl = StepLogger("advisor")
verdict = critic_output.get("verdict", "SKIPPED")
sl.log("INIT", "RUNNING", f"critic_verdict={verdict} | input_type={input_type}")
# (comment encoding corrupted)
advisor_input = dict(analyst_output)
if verdict == "DOWNGRADE":
advisor_input["_critic_note"] = (
f"Critic scored {critic_output.get('weighted_score', 0):.1f}/100. "
f"Challenges: {critic_output.get('challenges', [])}. "
"Use conservative risk assessment."
)
logger.info(
"Critic verdict=DOWNGRADE: Advisor will use conservative assessment"
)
t0 = time.time()
try:
result = run_advisor_pipeline(advisor_input)
duration_ms = int((time.time() - t0) * 1000)
risk = result.get("risk_score", 0)
urgent = len(result.get("actions", {}).get("urgent", []))
sl.log("COMPLETE", "SUCCESS", f"risk_score={risk} urgent={urgent}", duration_ms)
return result, sl
except Exception as e:
duration_ms = int((time.time() - t0) * 1000)
logger.error("Advisor Stage failed: %s", e)
sl.log("COMPLETE", "FAILED", str(e)[:100], duration_ms)
degradation_status.degrade("Advisor", str(e))
# (comment encoding corrupted)
# (comment encoding corrupted)
# (comment encoding corrupted)
urgent_actions = []
important_actions = []
code_patterns_fallback = [] # 脣 CODE-pattern 靘 UI
_SEVERITY_MAP = {"CRITICAL": "urgent", "HIGH": "important"}
for entry in advisor_input.get("analysis", []):
finding_id = entry.get("finding_id", "")
cve_id = entry.get("cve_id")
sev = entry.get("severity") or entry.get("adjusted_risk", "MEDIUM")
bucket = _SEVERITY_MAP.get(sev, "important")
# (comment encoding corrupted)
# (comment encoding corrupted)
if finding_id.startswith("CODE-") or entry.get("pattern_type"):
pt = entry.get("pattern_type", "UNKNOWN")
cwe = entry.get("cwe_id", "CWE-unknown")
snippet = entry.get("snippet", "")
code_patterns_fallback.append({
"finding_id": finding_id,
"cve_id": None,
"cwe_id": cwe,
"pattern_type": pt,
"package": f"Custom Code Pattern",
"severity": sev,
"action": f"Fix {pt} vulnerability ({cwe}). {entry.get('reasoning', '')}",
"vulnerable_snippet": snippet[:200],
"reason": entry.get("reasoning", f"Security Guard deterministic detection: {pt}"),
"is_repeated": False,
"_constitution_note": "CODE-pattern: not in URGENT per CI-1/CI-2. See code_patterns_summary.",
})
# (comment encoding corrupted)
elif cve_id and (str(cve_id).startswith("CVE-") or str(cve_id).startswith("GHSA-")):
action_entry = {
"cve_id": cve_id,
"package": entry.get("package", "unknown"),
"severity": sev,
"action": f"Review and patch {cve_id}.",
"command": f"# Investigate {cve_id}",
"reason": entry.get("reasoning", "Advisor degraded - manual review required."),
"is_repeated": False,
}
if bucket == "urgent":
urgent_actions.append(action_entry)
else:
important_actions.append(action_entry)
risk_score = max(
50,
min(100, len(urgent_actions) * 25 + len(important_actions) * 10),
)
summary_parts = []
if urgent_actions:
summary_parts.append(f"{len(urgent_actions)} package CVE(s) require immediate patching")
if code_patterns_fallback:
summary_parts.append(f"{len(code_patterns_fallback)} code-level pattern(s) detected (see Code Analysis tab)")
if important_actions:
summary_parts.append(f"{len(important_actions)} high-severity CVE(s) need review within 72h")
if not summary_parts:
summary_parts.append("System degraded. Please review raw scan data manually.")
logger.info(
"[DEGRADED] Advisor fallback: %d urgent CVEs + %d important CVEs + %d code-patterns (not in URGENT per CI-1/CI-2)",
len(urgent_actions), len(important_actions), len(code_patterns_fallback),
)
return {
"executive_summary": ". ".join(summary_parts),
"actions": {
"urgent": urgent_actions,
"important": important_actions,
"resolved": [],
},
"code_patterns_summary": code_patterns_fallback,
"risk_score": risk_score,
"risk_trend": "+0",
"scan_count": 1,
"generated_at": datetime.now(timezone.utc).isoformat(),
"_degraded": True,
"_error": str(e),
}, sl
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
def stage_orchestrator(
tech_stack: str,
feedback_from_judge: dict | None = None,
) -> tuple[Any, dict, StepLogger]:
"""
Stage 0: Orchestrator 瘙箏頝臬嚗A/B/C/D嚗
(OrchestrationContext, task_plan, StepLogger)
Graceful Degradation: 憭望唾楝敺 B嚗摰渡撘蝣潘雿箏券閮准
"""
from agents.orchestrator import run_orchestration
sl = StepLogger("orchestrator")
sl.log("INIT", "RUNNING", f"tech_stack={tech_stack[:60]}")
t0 = time.time()
try:
ctx, task_plan = run_orchestration(
user_input=tech_stack,
feedback_from_judge=feedback_from_judge,
)
task_plan = _normalize_pipeline_task_plan(task_plan)
duration_ms = int((time.time() - t0) * 1000)
scan_path = task_plan.get("path", "B")
sl.log("COMPLETE", "SUCCESS", f"path={scan_path}", duration_ms)
logger.info("[ORCH] Scan path: %s | plan: %s", scan_path, task_plan.get("agents_to_run", []))
return ctx, task_plan, sl
except Exception as e:
duration_ms = int((time.time() - t0) * 1000)
logger.error("Orchestrator Stage failed: %s", e)
sl.log("COMPLETE", "FAILED", str(e)[:100], duration_ms)
degradation_status.degrade("Orchestrator", str(e))
# (comment encoding corrupted)
from agents.orchestrator import OrchestrationContext, ScanPath
ctx = OrchestrationContext()
ctx.scan_path = ScanPath.FULL_CODE
task_plan = {
"path": "B",
"parallel_layer1": ["security_guard", "scout"],
"fusion_after_discovery": True,
"debate_cluster": True,
"judge": True,
"agents_to_run": ["security_guard", "scout", "intel_fusion", "analyst", "debate", "judge"],
"_degraded": True,
}
return ctx, task_plan, sl
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
def _build_code_patterns_summary(sg_result: dict) -> list[dict]:
"""
Security Guard patterns + hardcoded secrets を
CWE-enriched code_patterns_summary に変換する。
確定性抽出 + MITRE CWE v4.14 佐証注入。
LLM に依存しない。
"""
from tools.cwe_registry import build_cwe_reference, get_pattern_meta, severity_rank
def _attach_cwe_reference(entry: dict, cwe_id: str) -> dict:
"""將 MITRE CWE 官方定義注入 entry 的 cwe_reference 欄位"""
cwe_reference = build_cwe_reference(cwe_id)
if cwe_reference:
entry["cwe_reference"] = cwe_reference
return entry
# ── Step 1: 對 patterns 去重(同行 + 前 30 字元相同視為重複,保留嚴重性最高者)
def _normalized_line_no(item: dict) -> int:
"""Normalize scanner line metadata to the line_no contract."""
for key in ("line_no", "line", "source_line"):
value = item.get(key)
try:
line_no = int(value)
except (TypeError, ValueError):
continue
if line_no > 0:
return line_no
return 0
_dedup: dict[tuple, dict] = {}
for p in sg_result.get("patterns", []):
pt = p.get("pattern_type", "UNKNOWN")
sev = get_pattern_meta(pt).severity
key = (_normalized_line_no(p), str(p.get("snippet", ""))[:30])
existing = _dedup.get(key)
if existing is None:
_dedup[key] = p
else:
ex_pt = existing.get("pattern_type", "UNKNOWN")
ex_sev = get_pattern_meta(ex_pt).severity
if severity_rank(sev) > severity_rank(ex_sev):
_dedup[key] = p
code_patterns: list[dict] = []
counter = 1
# ── Step 2: 為每個 pattern 建立 finding entry + 注入 CWE reference
for p in _dedup.values():
pt = p.get("pattern_type", "UNKNOWN")
meta = get_pattern_meta(pt)
cwe = meta.cwe_id
owasp = meta.owasp_category
severity = meta.severity
entry = {
"finding_id": f"CODE-{counter:03d}",
"type": "code_pattern",
"pattern_type": pt,
"cwe_id": cwe,
"owasp_category": owasp,
"severity": severity,
"snippet": str(p.get("snippet", ""))[:200],
"line_no": _normalized_line_no(p),
"language": sg_result.get("language", "unknown"),
"canonical_cwe_id": cwe,
"weakness_family": meta.weakness_family,
"evidence_type": meta.evidence_type,
"coverage_level": p.get("coverage_level", "pattern"),
"confidence": p.get("confidence", "MEDIUM"),
}
entry["source_location"] = f"L{entry['line_no']}" if entry["line_no"] > 0 else "Line not provided by scanner"
entry = _attach_cwe_reference(entry, cwe)
code_patterns.append(entry)
counter += 1
# ── Step 3: 硬編碼密鑰
for h in sg_result.get("hardcoded", []):
entry = {
"finding_id": f"CODE-{counter:03d}",
"type": "hardcoded_secret",
"pattern_type": "HARDCODED_SECRET",
"cwe_id": "CWE-798",
"owasp_category": "A07:2021-Identification and Authentication Failures",
"severity": "HIGH",
"snippet": f"{h.get('name', 'secret')} = '****' (value redacted)",
"line_no": _normalized_line_no(h),
"language": sg_result.get("language", "unknown"),
"canonical_cwe_id": "CWE-798",
"weakness_family": "hardcoded_secret",
"evidence_type": "code_scan",
"coverage_level": h.get("coverage_level", "pattern"),
"confidence": h.get("confidence", "HIGH"),
}
entry["source_location"] = f"L{entry['line_no']}" if entry["line_no"] > 0 else "Line not provided by scanner"
entry = _attach_cwe_reference(entry, "CWE-798")
code_patterns.append(entry)
counter += 1
logger.info(
"[CODE_PATTERNS] Built %d CWE-enriched patterns (from %d raw patterns, %d hardcoded)",
len(code_patterns),
len(sg_result.get("patterns", [])),
len(sg_result.get("hardcoded", [])),
)
return code_patterns
def _summarize_vulnerabilities(vulnerabilities: list[dict[str, Any]]) -> dict[str, int]:
"""根據漏洞清單建立前端可直接使用的摘要數字。"""
summary = {
"total": 0,
"critical": 0,
"high": 0,
"medium": 0,
"low": 0,
"new": 0,
"new_since_last_scan": 0,
}
for vuln in vulnerabilities:
summary["total"] += 1
severity = str(vuln.get("severity", "LOW")).upper()
if severity == "CRITICAL":
summary["critical"] += 1
elif severity == "HIGH":
summary["high"] += 1
elif severity == "MEDIUM":
summary["medium"] += 1
else:
summary["low"] += 1
if vuln.get("is_new"):
summary["new"] += 1
summary["new_since_last_scan"] += 1
return summary
def _build_scan_report_payload(
scout_output: dict[str, Any],
advisor_output: dict[str, Any],
layer1_results: dict[str, Any],
) -> dict[str, Any]:
"""整理本次 scan 專屬的漏洞明細,避免 UI 回頭讀取全域記憶。"""
intel_applied = scout_output.get("_intel_fusion_applied", {})
vulnerabilities: list[dict[str, Any]] = []
for vuln in scout_output.get("vulnerabilities", []):
entry = dict(vuln)
entry.setdefault("source", "SCOUT")
enriched_by = list(entry.get("enriched_by", []))
if intel_applied and entry.get("source") != "INTEL_FUSION":
if any(
field in entry
for field in (
"composite_score",
"epss_score",
"ghsa_severity",
"otx_threat",
"intel_confidence",
"dimensions_used",
"weights_used",
"in_cisa_kev",
)
):
if "INTEL_FUSION" not in enriched_by:
enriched_by.append("INTEL_FUSION")
if enriched_by:
entry["enriched_by"] = enriched_by
vulnerabilities.append(entry)
detail_source = "scout_final_output"
if vulnerabilities:
summary = dict(scout_output.get("summary", {}))
summary.setdefault("new", summary.get("new_since_last_scan", 0))
summary.setdefault("new_since_last_scan", summary.get("new", 0))
else:
detail_source = "advisor_actions_fallback"
for level in ("urgent", "important", "resolved"):
for item in advisor_output.get("actions", {}).get(level, []):
vulnerabilities.append({
"cve_id": item.get("cve_id", ""),
"package": item.get("package", "unknown"),
"cvss_score": item.get("cvss_score", 0),
"severity": item.get("severity", "MEDIUM"),
"description": item.get("action", ""),
"is_new": item.get("is_new", False),
"source": "ADVISOR_ACTIONS",
"report_level": level.upper(),
})
summary = _summarize_vulnerabilities(vulnerabilities)
layer1_agents = [
name for name in ("security_guard", "scout", "intel_fusion")
if name in layer1_results
]
degraded_agents = [
name for name in layer1_agents
if isinstance(layer1_results.get(name), dict) and layer1_results[name].get("_degraded")
]
if not layer1_agents:
layer1_state = "skipped"
elif degraded_agents:
layer1_state = "degraded"
else:
layer1_state = "merged"
report_sources: dict[str, Any] = {
"vulnerability_detail": detail_source,
"enriched_by": ["intel_fusion"] if intel_applied else [],
"fallbacks": ["advisor_actions"] if detail_source == "advisor_actions_fallback" else [],
"layer1_state": layer1_state,
}
if degraded_agents:
report_sources["degraded_agents"] = degraded_agents
if intel_applied:
report_sources["intel_fusion_applied"] = intel_applied
if scout_output.get("representative_cve_evidence"):
report_sources["representative_cve_evidence_count"] = len(
scout_output.get("representative_cve_evidence", [])
)
return {
"vulnerability_detail": vulnerabilities,
"vulnerability_summary": summary,
"report_sources": report_sources,
"representative_cve_evidence": scout_output.get("representative_cve_evidence", []),
}
def _run_layer1_parallel(
tech_stack: str,
task_plan: dict,
on_progress: Any,
input_type: str = "pkg",
) -> dict[str, Any]:
"""
頝臬 B Layer 1 銝西瑁嚗雿輻 ThreadPoolExecutor 撠孵嚗
Why ThreadPoolExecutor嚗 asyncio嚗嚗
- main.py 臬甇亦撘蝣潘SSE callback 銋臬甇亦
- ThreadPoolExecutor 臬典甇亦啣銝銝西瑁憭 IO-bound 撌乩嚗LLM 澆恬
- 撠孵嚗銝閬瑽 callback 璈
銝西瑁撠鞊∴閬 task_plan 瘙箏嚗嚗
- security_guard嚗敺蝔撘蝣潭賢/憟隞/璅∪嚗頛詨伐tech_stack嚗
- intel_fusion嚗剔雁望亥岷嚗頛詨伐tech_stack嚗
Args:
tech_stack: 雿輻刻頛詨
task_plan: Orchestrator 隞餃閬
on_progress: SSE 脣漲隤
Returns:
{"security_guard": dict, "intel_fusion": dict}
"""
parallel_agents = task_plan.get("parallel_layer1", [])
layer1_results: dict[str, Any] = {}
def _run_security_guard() -> tuple[str, dict]:
""" Thread 銝剖瑁 Security Guard嚗 LLM 嚗"""
try:
from agents.security_guard import run_security_guard
result = run_security_guard(tech_stack, on_progress)
return ("security_guard", result)
except Exception as e:
logger.error("[LAYER1] Security Guard failed: %s", e)
degradation_status.degrade("Security Guard", str(e))
return ("security_guard", {
"extraction_status": "degraded",
"functions": [], "imports": [], "patterns": [], "hardcoded": [],
"stats": {"total_lines": 0, "functions_found": 0, "patterns_found": 0},
"_degraded": True, "_error": str(e),
})
def _run_intel_fusion(intel_input: list | str) -> tuple[str, dict]:
""" Thread 銝剖瑁 Intel Fusion嚗剔雁望亥岷嚗
v3.4 靽桀儔嚗亙憟隞嗅蝔勗銵剁憪蝔撘蝣潘嚗閫瘙 0 CVE 憿
intel_input 臭誑荔
- list[str]嚗憟隞嗅蝔勗銵剁 package_extractor 敺喳伐
- str嚗憪 tech_stack嚗fallback嚗Path A 憟隞嗆格芋撘嚗
"""
try:
from agents.intel_fusion import run_intel_fusion
result = run_intel_fusion(intel_input, on_progress)
return ("intel_fusion", result)
except Exception as e:
logger.error("[LAYER1] Intel Fusion failed: %s", e)
degradation_status.degrade("Intel Fusion", str(e))
return ("intel_fusion", {
"fusion_results": [],
"strategy_applied": "degraded",
"api_health_summary": {},
"_degraded": True, "_error": str(e),
})
def _run_scout() -> tuple[str, dict]:
"""Run Scout in parallel with Security Guard during discovery."""
try:
scout_target = tech_stack
scout_input_type = input_type
local_packages: list[str] = []
if input_type in {"code", "injection", "config"}:
from agents.security_guard import extract_code_surface
from tools.package_extractor import (
format_packages_for_intel_fusion,
packages_from_security_guard,
)
local_packages = packages_from_security_guard(extract_code_surface(tech_stack))
if not local_packages:
return ("scout", {
"scan_id": f"scan_code_only_{int(time.time())}",
"timestamp": datetime.now(timezone.utc).isoformat(),
"tech_stack": [],
"vulnerabilities": [],
"summary": {
"total": 0,
"new_since_last_scan": 0,
"critical": 0,
"high": 0,
"medium": 0,
"low": 0,
},
"_duration_ms": 0,
"_package_scan_skipped": True,
"_skip_reason": "source_code_without_third_party_packages",
"_extracted_packages_local": [],
})
scout_target = format_packages_for_intel_fusion(local_packages)
scout_input_type = "pkg"
result, scout_sl = stage_scout(scout_target, input_type=scout_input_type)
if scout_sl.steps:
result["_duration_ms"] = scout_sl.steps[-1].get("duration_ms", 0)
if local_packages:
result["_extracted_packages_local"] = local_packages
result["_scout_target_input"] = scout_target
result["_scout_input_type"] = scout_input_type
return ("scout", result)
except Exception as e:
logger.error("[LAYER1] Scout failed: %s", e)
degradation_status.degrade("Scout", str(e))
return ("scout", {
"scan_id": f"scan_degraded_{int(time.time())}",
"vulnerabilities": [],
"summary": {"total": 0, "critical": 0, "high": 0, "medium": 0, "low": 0},
"_degraded": True, "_error": str(e),
})
def _extract_packages_from_sg_result(sg_result: dict[str, Any]) -> list[str]:
"""從 Security Guard 結果萃取第三方套件,供後續 Scout 使用。"""
try:
from tools.package_extractor import packages_from_security_guard
extracted = packages_from_security_guard(sg_result)
logger.info(
"[LAYER1] PackageExtractor: %d packages extracted from imports: %s",
len(extracted), extracted,
)
return extracted
except Exception as pe:
logger.warning("[LAYER1] PackageExtractor failed (fallback to raw input): %s", pe)
return []
def _extract_cwe_targets_from_sg_result(sg_result: dict[str, Any]) -> list[str]:
"""從 Security Guard patterns 中整理 CWE targets,供 observability 與 fallback 使用。"""
sg_patterns = sg_result.get("patterns", [])
seen_cwe: set[str] = set()
cwe_targets: list[str] = []
for pattern in sg_patterns:
cwe = pattern.get("cwe_id", "")
if cwe and cwe.startswith("CWE-") and cwe not in seen_cwe:
seen_cwe.add(cwe)
cwe_targets.append(cwe)
return cwe_targets
extracted_packages: list[str] = []
cwe_targets: list[str] = []
requested_workers = [
agent_name
for agent_name in ("security_guard", "scout")
if agent_name in parallel_agents
]
if not requested_workers:
return layer1_results
futures: dict[Any, str] = {}
max_workers = len(requested_workers)
with ThreadPoolExecutor(max_workers=max_workers, thread_name_prefix="layer1") as executor:
if "security_guard" in requested_workers:
rate_limiter.wait_if_needed("security_guard")
futures[executor.submit(_run_security_guard)] = "security_guard"
if "scout" in requested_workers:
rate_limiter.wait_if_needed("scout")
if on_progress:
on_progress("scout", "RUNNING", {"step": "discovery_parallel"})
futures[executor.submit(_run_scout)] = "scout"
for future in as_completed(futures):
agent_name = futures[future]
try:
result_name, result_payload = future.result()
except Exception as exc:
logger.error("[LAYER1] %s future failed unexpectedly: %s", agent_name, exc)
degradation_status.degrade(agent_name, str(exc))
continue
layer1_results[result_name] = result_payload
logger.info("[LAYER1] %s complete", result_name)
if result_name == "security_guard":
sg_extracted = _extract_packages_from_sg_result(result_payload)
if sg_extracted or not extracted_packages:
extracted_packages = sg_extracted
cwe_targets = _extract_cwe_targets_from_sg_result(result_payload)
if cwe_targets:
logger.info(
"[LAYER1] Security Guard produced %d CWE targets: %s",
len(cwe_targets), cwe_targets,
)
elif result_name == "scout" and result_payload.get("_extracted_packages_local") and not extracted_packages:
extracted_packages = list(result_payload.get("_extracted_packages_local") or [])
layer1_results["_extracted_packages"] = extracted_packages
layer1_results["_cwe_targets"] = cwe_targets
return layer1_results
def _build_intel_fusion_input(
scout_output: dict[str, Any],
sg_result: dict[str, Any] | None,
code_patterns: list[dict[str, Any]],
extracted_packages: list[str],
tech_stack: str,
input_type: str,
) -> dict[str, Any]:
"""Build a clean post-discovery input for Intel Fusion."""
cve_ids: list[str] = []
seen_cves: set[str] = set()
for vuln in scout_output.get("vulnerabilities", []):
cve_id = str(vuln.get("cve_id", "")).strip()
if (cve_id.startswith("CVE-") or cve_id.startswith("GHSA-")) and cve_id not in seen_cves:
seen_cves.add(cve_id)
cve_ids.append(cve_id)
cwe_ids: list[str] = []
seen_cwes: set[str] = set()
for pattern in code_patterns:
cwe_id = str(pattern.get("cwe_id", "")).strip()
if cwe_id.startswith("CWE-") and cwe_id not in seen_cwes:
seen_cwes.add(cwe_id)
cwe_ids.append(cwe_id)
return {
"mode": "post_discovery",
"input_type": input_type,
"cve_ids": cve_ids,
"cwe_ids": cwe_ids,
"packages": extracted_packages,
"scout_vulnerabilities": scout_output.get("vulnerabilities", [])[:20],
"code_patterns": code_patterns[:20],
"security_guard_stats": (sg_result or {}).get("stats", {}),
"raw_input_fingerprint": str(hash(tech_stack)),
}
def _merge_intel_fusion_into_scout(
scout_output: dict[str, Any],
intel_fusion_result: dict[str, Any],
) -> dict[str, Any]:
"""Merge Intel Fusion evidence after Scout completes discovery."""
if not intel_fusion_result:
return scout_output
try:
from agents.scout import merge_intel_fusion_evidence
return merge_intel_fusion_evidence(scout_output, intel_fusion_result)
except Exception as exc:
logger.warning("[PIPELINE] Intel Fusion merge skipped: %s", exc)
scout_output["intel_fusion_result"] = intel_fusion_result
return scout_output
def run_pipeline(tech_stack: str, input_type: str = "pkg") -> dict[str, Any]:
"""
瑁摰渡 ThreatHunter 蝞∠嚗v3.1 Orchestrator 撽嚗
Pipeline: Orchestrator [Layer 1 銝西] Scout Analyst [Critic] Advisor
v3.7: input_type 瘙箏 Path-Aware Skills 頝舐晞
pkg=憟隞嗆 / code=皞蝣澆祟閮 / injection=AI摰 / config=閮剖瑼
v3.9 Sandbox: SANDBOX_ENABLED=true 銝 Docker 舐剁典捆典批瑁
Args:
tech_stack: 雿輻刻頛詨亦銵摮銝莎憒 "Django 4.2, Redis 7.0"嚗
input_type: 蝡臬菜葫頛詨仿 (pkg/code/config/injection)
Returns:
怠 Advisor 銵勗 dict嚗銝 pipeline_meta 甈雿
"""
# (comment encoding corrupted)
if SANDBOX_ENABLED and _DOCKER_SANDBOX_OK and is_docker_available():
logger.info("[SANDBOX] Docker isolation ACTIVE delegating to container")
result = run_in_sandbox(tech_stack=tech_stack, input_type=input_type)
if not result.get("fallback"):
return result # 摰孵典瑁
# (comment encoding corrupted)
logger.warning(
"[SANDBOX] Container fallback: %s using in-process mode",
result.get("error", "unknown"),
)
# (comment encoding corrupted)
return run_pipeline_with_callback(tech_stack, progress_callback=None, input_type=input_type)
def run_pipeline_sync(tech_stack: str, input_type: str = "pkg") -> dict[str, Any]:
"""
run_pipeline 甇亙亙嚗靘 sandbox/sandbox_runner.py 典捆典批澆恬
瘜冽嚗摰孵典批澆急迨賢 SANDBOX_ENABLED=false嚗踹餈湧脣 Docker
"""
# (comment encoding corrupted)
return run_pipeline_with_callback(tech_stack, progress_callback=None, input_type=input_type)
def _run_pipeline_with_callback_legacy_v31(
tech_stack: str,
progress_callback: Any = None,
input_type: str = "pkg",
) -> dict[str, Any]:
"""
瑁摰 Pipeline嚗瘥 Stage 摰敺澆 progress_callback
v3.7: input_type 瘙箏 Agent 頛芸 Skill SOP嚗Path-Aware Skills嚗
Args:
tech_stack: 銵摮銝
progress_callback: 交 (agent_name: str, status: str, detail: dict) 賢
input_type: 頛詨仿 (pkg / code / config / injection)
Returns:
怠 Advisor 銵勗 dict
"""
pipeline_start = time.time()
completed_stages: list[str] = []
stages_detail: dict[str, Any] = {}
# (comment encoding corrupted)
orch_ctx: Any = None
task_plan: dict = {"path": "B"} # 閮剛楝敺 B
layer1_results: dict[str, Any] = {}
feedback_loop_count: int = 0
MAX_FEEDBACK_LOOPS: int = 2
def _notify(agent: str, status: str, detail: dict) -> None:
if progress_callback:
try:
progress_callback(agent, status, detail)
except Exception:
pass # callback 憭望銝敶梢 pipeline
logger.info("=" * 60)
logger.info(" ThreatHunter Pipeline v3.1 Start (Orchestrator-driven)")
logger.info(" tech_stack : %s", tech_stack)
logger.info(" input_type : %s", input_type)
logger.info(" ENABLE_CRITIC: %s", ENABLE_CRITIC)
logger.info("=" * 60)
# (comment encoding corrupted)
from checkpoint import recorder
recorder.start_scan(f"pipe_{int(pipeline_start)}")
# (comment encoding corrupted)
# (comment encoding corrupted)
l0_report: dict[str, Any] = {}
sanitized_stack: str = tech_stack
try:
from input_sanitizer import sanitize_input, format_l0_report
san_result = sanitize_input(tech_stack)
l0_report = format_l0_report(san_result)
if not san_result.safe:
# (comment encoding corrupted)
logger.warning("[L0] Input BLOCKED: %s", san_result.blocked_reason)
_notify("input_sanitizer", "COMPLETE", {
"status": "BLOCKED",
"reason": san_result.blocked_reason,
})
return {
"executive_summary": f"頛詨亥◤摰券瞈曉冽蝯嚗{san_result.blocked_reason}",
"blocked": True,
"actions": {"urgent": [], "important": [], "resolved": []},
"risk_score": 0,
"risk_trend": "+0",
"pipeline_meta": {
"pipeline_version": "3.1",
"tech_stack": tech_stack,
"stages_completed": 0,
"stages_detail": {},
"enable_critic": ENABLE_CRITIC,
"critic_verdict": "SKIPPED",
"critic_score": 0.0,
"duration_seconds": 0.0,
"degradation": {"level": 5, "label": "INPUT_BLOCKED"},
"generated_at": datetime.now(timezone.utc).isoformat(),
"l0_report": l0_report,
},
}
sanitized_stack = san_result.sanitized_input
if san_result.truncated:
logger.warning("[L0] Input truncated: %d %d chars",
san_result.original_length, len(sanitized_stack))
logger.info("[L0] OK: type=%s l0_findings=%d hash=%s",
san_result.input_type, len(san_result.l0_findings), san_result.input_hash)
_notify("input_sanitizer", "COMPLETE", {
"status": "SUCCESS",
"input_type": san_result.input_type,
"l0_warning_count": l0_report.get("l0_warning_count", 0),
"truncated": san_result.truncated,
})
recorder.stage_exit("input_sanitizer", "SUCCESS", {
"input_type": san_result.input_type,
"l0_warning_count": l0_report.get("l0_warning_count", 0),
}, 0)
except ImportError:
# (comment encoding corrupted)
logger.warning("[L0] input_sanitizer not available, skipping L0 filter")
except Exception as e:
# (comment encoding corrupted)
logger.error("[L0] Sanitizer error (non-fatal): %s", e)
# (comment encoding corrupted)
_notify("orchestrator", "RUNNING", {})
orch_ctx, task_plan, orch_sl = stage_orchestrator(sanitized_stack)
scan_path = task_plan.get("path", "B")
orch_detail = {
"status": "SUCCESS" if not task_plan.get("_degraded") else "DEGRADED",
"scan_path": scan_path,
"agents_to_run": task_plan.get("agents_to_run", []),
"duration_ms": orch_sl.steps[-1].get("duration_ms", 0) if orch_sl.steps else 0,
"l0_input_type": l0_report.get("input_type", "unknown"),
}
stages_detail["orchestrator"] = orch_detail
completed_stages.append("orchestrator")
_notify("orchestrator", "COMPLETE", orch_detail)
recorder.stage_enter("orchestrator", {"tech_stack": sanitized_stack[:200]})
recorder.stage_exit("orchestrator", orch_detail.get("status", "SUCCESS"), {
"scan_path": scan_path,
"agents_to_run": task_plan.get("agents_to_run", []),
}, orch_detail.get("duration_ms", 0))
logger.info("[PIPELINE] Scan path: %s", scan_path)
# (comment encoding corrupted)
# (comment encoding corrupted)
# (comment encoding corrupted)
# (comment encoding corrupted)
extracted_packages: list[str] = []
if scan_path in ("A", "B"):
parallel_agents = task_plan.get("parallel_layer1", [])
if parallel_agents:
_notify("layer1_parallel", "RUNNING", {"agents": parallel_agents})
layer1_results = _run_layer1_parallel(tech_stack, task_plan, _notify)
_notify("layer1_parallel", "COMPLETE", {
"agents_completed": list(layer1_results.keys()),
})
# (comment encoding corrupted)
if "intel_fusion" in layer1_results:
extracted_packages = layer1_results["intel_fusion"].get("_extracted_packages", [])
if extracted_packages:
logger.info(
"[PIPELINE] extracted_packages from layer1: %s", extracted_packages
)
# (comment encoding corrupted)
for agent_name, result in layer1_results.items():
is_degraded = result.get("_degraded", False)
agent_detail = {
"status": "DEGRADED" if is_degraded else "SUCCESS",
"duration_ms": result.get("_duration_ms", 0),
# (comment encoding corrupted)
"_degraded": is_degraded,
"_error": result.get("_error", "") if is_degraded else "",
}
# (comment encoding corrupted)
if agent_name == "security_guard":
agent_detail["functions_found"] = result.get("stats", {}).get("functions_found", 0)
agent_detail["patterns_found"] = result.get("stats", {}).get("patterns_found", 0)
agent_detail["injection_detected"] = result.get("injection_attempts_detected", False)
# (comment encoding corrupted)
elif agent_name == "intel_fusion":
agent_detail["cves_scored"] = len(result.get("fusion_results", []))
stages_detail[agent_name] = agent_detail
completed_stages.append(agent_name)
recorder.stage_exit(agent_name, agent_detail["status"], agent_detail, agent_detail.get("duration_ms", 0))
# (comment encoding corrupted)
if orch_ctx is not None:
for agent_name, result in layer1_results.items():
try:
orch_ctx.store_result(agent_name, result)
except Exception:
pass
# (comment encoding corrupted)
# (comment encoding corrupted)
_notify("scout", "RUNNING", {})
rate_limiter.wait_if_needed("scout")
# (comment encoding corrupted)
# (comment encoding corrupted)
scout_input: str
if extracted_packages:
from tools.package_extractor import format_packages_for_intel_fusion
scout_input = format_packages_for_intel_fusion(extracted_packages)
logger.info("[PIPELINE] Scout using extracted packages: %s", scout_input)
else:
scout_input = tech_stack
logger.info("[PIPELINE] Scout using raw tech_stack (no packages extracted)")
scout_output, scout_sl = stage_scout(
scout_input,
input_type=input_type,
intel_fusion_result=layer1_results.get("intel_fusion"),
)
# 保留原始 Layer 1 情報,方便 UI 與除錯觀察。
if "intel_fusion" in layer1_results:
scout_output["intel_fusion_result"] = layer1_results["intel_fusion"]
# (comment encoding corrupted)
# (comment encoding corrupted)
# (comment encoding corrupted)
# (comment encoding corrupted)
# (comment encoding corrupted)
_sg_code_patterns: list[dict] = [] # 函式級別變數,確保 final return 可存取
if "security_guard" in layer1_results:
_sg_code_patterns = _build_code_patterns_summary(
layer1_results["security_guard"]
)
if _sg_code_patterns:
scout_output["code_patterns"] = _sg_code_patterns
logger.info(
"[PIPELINE] v4.0: SG code_patterns injected into scout_output "
"(%d patterns, Path=%s) — will be merged into final result",
len(_sg_code_patterns),
input_type,
)
else:
logger.debug("[PIPELINE] v4.0: SG returned no code_patterns (clean code path)")
scout_detail = {
"status": "SUCCESS" if not scout_output.get("_degraded") else "DEGRADED",
"vuln_count": len(scout_output.get("vulnerabilities", [])),
"duration_ms": scout_sl.steps[-1].get("duration_ms", 0) if scout_sl.steps else 0,
"packages_used": extracted_packages,
}
stages_detail["scout"] = scout_detail
completed_stages.append("scout")
_notify("scout", "COMPLETE", scout_detail)
# (comment encoding corrupted)
from agents.scout import SKILL_MAP as SCOUT_SKILL_MAP
recorder.stage_enter("scout", {"tech_stack": scout_input[:200], "packages": extracted_packages},
skill_file=SCOUT_SKILL_MAP.get(input_type, "threat_intel.md"),
input_type=input_type)
recorder.stage_exit("scout", scout_detail.get("status", "SUCCESS"), scout_output, scout_detail.get("duration_ms", 0))
# (comment encoding corrupted)
if scan_path == "C":
logger.info("[PIPELINE] Path C: skipping Analyst + Critic direct to Advisor")
_notify("analyst", "COMPLETE", {"status": "SKIPPED", "reason": "path_C"})
_notify("critic", "COMPLETE", {"status": "SKIPPED", "reason": "path_C"})
# (comment encoding corrupted)
analyst_output: dict[str, Any] = {
"scan_id": scout_output.get("scan_id", "unknown"),
"risk_score": 30,
"risk_trend": "+0",
"analysis": [],
"_skipped": True,
"_reason": "path_C_doc_scan",
}
critic_output: dict[str, Any] = {
"verdict": "SKIPPED",
"weighted_score": 60.0,
"_skipped": True,
}
else:
# (comment encoding corrupted)
_notify("analyst", "RUNNING", {})
rate_limiter.wait_if_needed("analyst")
analyst_output, analyst_sl = stage_analyst(scout_output, input_type=input_type)
analyst_detail = {
"status": "SUCCESS" if not analyst_output.get("_degraded") else "DEGRADED",
"risk_score": analyst_output.get("risk_score", 0),
"duration_ms": analyst_sl.steps[-1].get("duration_ms", 0) if analyst_sl.steps else 0,
}
stages_detail["analyst"] = analyst_detail
completed_stages.append("analyst")
_notify("analyst", "COMPLETE", analyst_detail)
# v3.7: stage_enter with skill_file + input_type
from agents.analyst import SKILL_MAP as ANALYST_SKILL_MAP
recorder.stage_enter("analyst", scout_output,
skill_file=ANALYST_SKILL_MAP.get(input_type, "chain_analysis.md"),
input_type=input_type)
recorder.stage_exit("analyst", analyst_detail.get("status", "SUCCESS"), analyst_output, analyst_detail.get("duration_ms", 0))
# (comment encoding corrupted)
_notify("critic", "RUNNING", {})
rate_limiter.wait_if_needed("critic")
critic_output, critic_sl = stage_critic(analyst_output, input_type=input_type)
critic_detail = {
"status": "SUCCESS" if not critic_output.get("_degraded") else "DEGRADED",
"verdict": critic_output.get("verdict", "SKIPPED"),
"score": critic_output.get("weighted_score", 0),
"duration_ms": critic_sl.steps[-1].get("duration_ms", 0) if critic_sl.steps else 0,
}
stages_detail["critic"] = critic_detail
completed_stages.append("critic")
_notify("critic", "COMPLETE", critic_detail)
# v3.7: stage_enter with skill_file + input_type
from agents.critic import SKILL_MAP as CRITIC_SKILL_MAP
recorder.stage_enter("critic", analyst_output,
skill_file=CRITIC_SKILL_MAP.get(input_type, "debate_sop.md"),
input_type=input_type)
recorder.stage_exit("critic", critic_detail.get("status", "SUCCESS"), critic_output, critic_detail.get("duration_ms", 0))
# (comment encoding corrupted)
_notify("advisor", "RUNNING", {})
rate_limiter.wait_if_needed("advisor")
advisor_output, advisor_sl = stage_advisor(analyst_output, critic_output, input_type=input_type)
advisor_detail = {
"status": "SUCCESS" if not advisor_output.get("_degraded") else "DEGRADED",
"urgent_count": len(advisor_output.get("actions", {}).get("urgent", [])),
"duration_ms": advisor_sl.steps[-1].get("duration_ms", 0) if advisor_sl.steps else 0,
}
stages_detail["advisor"] = advisor_detail
completed_stages.append("advisor")
_notify("advisor", "COMPLETE", advisor_detail)
# v3.7: stage_enter with skill_file + input_type
from agents.advisor import SKILL_MAP as ADVISOR_SKILL_MAP
recorder.stage_enter("advisor", analyst_output,
skill_file=ADVISOR_SKILL_MAP.get(input_type, "action_report.md"),
input_type=input_type)
recorder.stage_exit("advisor", advisor_detail.get("status", "SUCCESS"), advisor_output, advisor_detail.get("duration_ms", 0))
# (comment encoding corrupted)
# (comment encoding corrupted)
advisor_confidence = advisor_output.get("confidence", "HIGH")
feedback_triggered = (
advisor_confidence in ("NEEDS_VERIFICATION", "LOW", "MEDIUM")
and feedback_loop_count < MAX_FEEDBACK_LOOPS
and scan_path != "D" # 脫迫頝臬 D ⊿擖
)
if feedback_triggered:
feedback_loop_count += 1
logger.warning(
"[PIPELINE] Advisor confidence=%s triggering Feedback Loop %d/%d",
advisor_confidence, feedback_loop_count, MAX_FEEDBACK_LOOPS,
)
_notify("feedback_loop", "RUNNING", {
"loop": feedback_loop_count,
"reason": f"confidence={advisor_confidence}",
})
# (comment encoding corrupted)
low_conf_cves = [
a.get("cve_id") for a in analyst_output.get("analysis", [])
if a.get("chain_risk", {}).get("confidence") in ("NEEDS_VERIFICATION", "LOW")
]
logger.info("[PIPELINE] Feedback Loop targeting CVEs: %s", low_conf_cves)
# (comment encoding corrupted)
if orch_ctx is not None:
try:
orch_ctx.feedback_loops = feedback_loop_count
except Exception:
pass
_notify("feedback_loop", "COMPLETE", {"loop": feedback_loop_count, "cves": low_conf_cves})
# (comment encoding corrupted)
orch_summary: dict = {}
if orch_ctx is not None:
try:
from agents.orchestrator import finalize_orchestration
orch_ctx.final_confidence = advisor_confidence
orch_summary = finalize_orchestration(orch_ctx)
except Exception as e:
logger.warning("[PIPELINE] finalize_orchestration failed: %s", e)
# (comment encoding corrupted)
duration = round(time.time() - pipeline_start, 2)
pipeline_meta = {
"pipeline_version": "3.1",
"tech_stack": tech_stack,
"scan_path": scan_path,
"stages_completed": len(completed_stages),
"stages_detail": stages_detail,
"enable_critic": ENABLE_CRITIC,
"critic_verdict": critic_output.get("verdict", "SKIPPED"),
"critic_score": critic_output.get("weighted_score", 0),
"feedback_loops": feedback_loop_count,
"duration_seconds": duration,
"degradation": degradation_status.to_dict(),
"orchestration": orch_summary,
"layer1_agents": list(layer1_results.keys()),
"generated_at": datetime.now(timezone.utc).isoformat(),
}
logger.info(
"Pipeline v3.1 COMPLETE in %.1fs | path=%s | feedback_loops=%d",
duration, scan_path, feedback_loop_count,
)
recorder.end_scan("COMPLETE", duration)
# ── Harness Layer 7: 強制注入 Security Guard CWE 佐證 ──────────────────
# _sg_code_patterns 由 _build_code_patterns_summary() 生成,含 MITRE CWE 定義。
# Advisor LLM 從不接收 code_patterns,因此必須在 return 前直接合併。
# 確保無論掃描路徑(A/B/C)都能在 API response 中看到 CWE 佐證。
report_payload = _build_scan_report_payload(scout_output, advisor_output, layer1_results)
final_output: dict[str, Any] = {
**advisor_output,
**report_payload,
"pipeline_meta": pipeline_meta,
}
if _sg_code_patterns:
existing_cps = final_output.get("code_patterns_summary", [])
final_output["code_patterns_summary"] = existing_cps + _sg_code_patterns
logger.info(
"[PIPELINE] Harness L7: %d CWE-enriched code patterns merged into final result",
len(_sg_code_patterns),
)
return final_output
def run_pipeline_with_callback(
tech_stack: str,
progress_callback: Any = None,
input_type: str = "pkg",
) -> dict[str, Any]:
"""
Canonical pipeline DAG:
Orchestrator -> (Security Guard + Scout in parallel) -> Intel Fusion
-> Analyst -> Critic -> Advisor.
"""
pipeline_start = time.time()
completed_stages: list[str] = []
stages_detail: dict[str, Any] = {}
layer1_results: dict[str, Any] = {}
def _notify(agent: str, status: str, detail: dict) -> None:
if not progress_callback:
return
try:
progress_callback(agent, status, detail)
except Exception as exc:
logger.debug("[PIPELINE] progress callback ignored: %s", exc)
def _remember_stage(agent: str, detail: dict[str, Any]) -> None:
stages_detail[agent] = detail
if agent not in completed_stages:
completed_stages.append(agent)
from checkpoint import recorder
recorder.start_scan(f"pipe_{int(pipeline_start)}")
l0_report: dict[str, Any] = {}
sanitized_stack = tech_stack
try:
from input_sanitizer import sanitize_input, format_l0_report
san_result = sanitize_input(tech_stack)
l0_report = format_l0_report(san_result)
if not san_result.safe:
_notify("input_sanitizer", "COMPLETE", {
"status": "BLOCKED",
"reason": san_result.blocked_reason,
})
duration = round(time.time() - pipeline_start, 2)
return {
"executive_summary": f"Input blocked: {san_result.blocked_reason}",
"blocked": True,
"actions": {"urgent": [], "important": [], "resolved": []},
"risk_score": 0,
"risk_trend": "+0",
"pipeline_meta": {
"pipeline_version": "4.0",
"tech_stack": tech_stack,
"scan_path": "BLOCKED",
"stages_completed": 0,
"stages_detail": {},
"enable_critic": ENABLE_CRITIC,
"critic_verdict": "SKIPPED",
"critic_score": 0.0,
"duration_seconds": duration,
"degradation": {"level": 5, "label": "INPUT_BLOCKED"},
"generated_at": datetime.now(timezone.utc).isoformat(),
"l0_report": l0_report,
},
}
sanitized_stack = san_result.sanitized_input
input_type = _canonical_runtime_input_type(input_type, san_result.input_type)
l0_detail = {
"status": "SUCCESS",
"input_type": san_result.input_type,
"pipeline_input_type": input_type,
"l0_warning_count": l0_report.get("l0_warning_count", 0),
"truncated": san_result.truncated,
}
_remember_stage("input_sanitizer", l0_detail)
_notify("input_sanitizer", "COMPLETE", l0_detail)
recorder.stage_exit("input_sanitizer", "SUCCESS", l0_detail, 0)
except Exception as exc:
logger.warning("[L0] Sanitizer skipped: %s", exc)
if input_type == "sql_review":
from tools.sql_syntax_reviewer import review_sql_syntax
_notify("sql_review", "RUNNING", {})
review = review_sql_syntax(sanitized_stack)
sql_detail = {
"status": "SUCCESS",
"patterns_detected": review.get("summary", {}).get("patterns_detected", 0),
"findings_count": review.get("summary", {}).get("total", 0),
"requires_application_context": True,
}
_remember_stage("sql_review", sql_detail)
_notify("sql_review", "COMPLETE", sql_detail)
recorder.stage_exit("sql_review", "SUCCESS", sql_detail, 0)
duration = round(time.time() - pipeline_start, 2)
recorder.end_scan("COMPLETE", duration)
return {
"executive_summary": (
"SQL syntax review completed. Findings are payload/syntax evidence only; "
"application source context is required before calling this a verified vulnerability."
),
"actions": {"urgent": [], "important": [], "resolved": []},
"risk_score": 0,
"risk_trend": "+0",
"sql_syntax_review": review,
"vulnerability_detail": [],
"vulnerability_summary": {
"total": 0,
"critical": 0,
"high": 0,
"medium": 0,
"low": 0,
"new": 0,
"new_since_last_scan": 0,
},
"report_sources": {
"vulnerability_detail": "sql_syntax_review",
"layer1_state": "skipped",
"external_scan_skipped": True,
},
"code_patterns_summary": [],
"pipeline_meta": {
"pipeline_version": "4.0",
"tech_stack": tech_stack,
"scan_path": "C",
"stages_completed": len(completed_stages),
"stages_detail": stages_detail,
"enable_critic": ENABLE_CRITIC,
"critic_verdict": "SKIPPED",
"critic_score": 0.0,
"feedback_loops": 0,
"duration_seconds": duration,
"degradation": degradation_status.to_dict(),
"orchestration": {},
"layer1_agents": [],
"fusion_after_discovery": False,
"l0_report": l0_report,
"generated_at": datetime.now(timezone.utc).isoformat(),
},
}
_notify("orchestrator", "RUNNING", {})
orch_ctx, task_plan, orch_sl = stage_orchestrator(sanitized_stack)
task_plan = _constrain_task_plan_by_input_type(task_plan, input_type)
scan_path = task_plan.get("path", "B")
orch_detail = {
"status": "SUCCESS" if not task_plan.get("_degraded") else "DEGRADED",
"scan_path": scan_path,
"agents_to_run": task_plan.get("agents_to_run", []),
"duration_ms": orch_sl.steps[-1].get("duration_ms", 0) if orch_sl.steps else 0,
"l0_input_type": l0_report.get("input_type", "unknown"),
"pipeline_input_type": input_type,
"route_corrected": task_plan.get("_route_corrected"),
}
_remember_stage("orchestrator", orch_detail)
_notify("orchestrator", "COMPLETE", orch_detail)
recorder.stage_enter("orchestrator", {"tech_stack": sanitized_stack[:200]})
recorder.stage_exit("orchestrator", orch_detail["status"], orch_detail, orch_detail["duration_ms"])
extracted_packages: list[str] = []
scout_input = sanitized_stack
scout_output: dict[str, Any] = {}
scout_duration_ms = 0
sg_code_patterns: list[dict[str, Any]] = []
intel_fusion_result: dict[str, Any] = {}
benchmark_context: dict[str, Any] | None = None
parallel_agents = task_plan.get("parallel_layer1", [])
if parallel_agents:
_notify("layer1_parallel", "RUNNING", {"agents": parallel_agents})
layer1_results = _run_layer1_parallel(
sanitized_stack, task_plan, _notify, input_type=input_type
)
visible_agents = [name for name in layer1_results if not name.startswith("_")]
_notify("layer1_parallel", "COMPLETE", {"agents_completed": visible_agents})
extracted_packages = layer1_results.get("_extracted_packages", [])
if "security_guard" in layer1_results:
sg_result = layer1_results["security_guard"]
sg_code_patterns = _build_code_patterns_summary(sg_result)
sg_detail = {
"status": "SUCCESS" if not sg_result.get("_degraded") else "DEGRADED",
"duration_ms": sg_result.get("_duration_ms", 0),
"functions_found": sg_result.get("stats", {}).get("functions_found", 0),
"patterns_found": sg_result.get("stats", {}).get("patterns_found", 0),
"injection_detected": sg_result.get("injection_attempts_detected", False),
"_degraded": sg_result.get("_degraded", False),
"_error": sg_result.get("_error", "") if sg_result.get("_degraded") else "",
}
_remember_stage("security_guard", sg_detail)
recorder.stage_exit("security_guard", sg_detail["status"], sg_detail, sg_detail["duration_ms"])
if "scout" in layer1_results:
scout_output = layer1_results["scout"]
scout_duration_ms = scout_output.get("_duration_ms", 0)
scout_detail = {
"status": "SUCCESS" if not scout_output.get("_degraded") else "DEGRADED",
"vuln_count": len(scout_output.get("vulnerabilities", [])),
"duration_ms": scout_duration_ms,
"packages_used": extracted_packages,
"_degraded": scout_output.get("_degraded", False),
"_error": scout_output.get("_error", "") if scout_output.get("_degraded") else "",
}
_remember_stage("scout", scout_detail)
_notify("scout", "COMPLETE", scout_detail)
if not scout_output:
_notify("scout", "RUNNING", {})
rate_limiter.wait_if_needed("scout")
scout_output, scout_sl = stage_scout(scout_input, input_type=input_type)
scout_duration_ms = scout_sl.steps[-1].get("duration_ms", 0) if scout_sl.steps else 0
scout_detail = {
"status": "SUCCESS" if not scout_output.get("_degraded") else "DEGRADED",
"vuln_count": len(scout_output.get("vulnerabilities", [])),
"duration_ms": scout_duration_ms,
"packages_used": extracted_packages,
}
_remember_stage("scout", scout_detail)
_notify("scout", "COMPLETE", scout_detail)
if sg_code_patterns:
scout_output["code_patterns"] = sg_code_patterns
from agents.scout import SKILL_MAP as SCOUT_SKILL_MAP
recorder.stage_enter("scout", {"tech_stack": scout_input[:200], "packages": extracted_packages},
skill_file=SCOUT_SKILL_MAP.get(input_type, "threat_intel.md"),
input_type=input_type)
recorder.stage_exit("scout", stages_detail["scout"]["status"], scout_output, scout_duration_ms)
if task_plan.get("fusion_after_discovery", False):
intel_input = _build_intel_fusion_input(
scout_output=scout_output,
sg_result=layer1_results.get("security_guard"),
code_patterns=sg_code_patterns,
extracted_packages=extracted_packages,
tech_stack=sanitized_stack,
input_type=input_type,
)
intel_fusion_result, intel_sl = stage_intel_fusion(intel_input, _notify)
layer1_results["intel_fusion"] = intel_fusion_result
scout_output["intel_fusion_result"] = intel_fusion_result
scout_output = _merge_intel_fusion_into_scout(scout_output, intel_fusion_result)
intel_detail = {
"status": "SUCCESS" if not intel_fusion_result.get("_degraded") else "DEGRADED",
"cves_scored": len(intel_fusion_result.get("fusion_results", [])),
"duration_ms": intel_sl.steps[-1].get("duration_ms", 0) if intel_sl.steps else 0,
"_degraded": intel_fusion_result.get("_degraded", False),
"_error": intel_fusion_result.get("_error", "") if intel_fusion_result.get("_degraded") else "",
}
_remember_stage("intel_fusion", intel_detail)
if orch_ctx is not None:
try:
orch_ctx.store_result("intel_fusion", intel_fusion_result)
except Exception as exc:
logger.debug("[PIPELINE] Orchestration store skipped for Intel Fusion: %s", exc)
if scan_path == "C":
analyst_output = {
"scan_id": scout_output.get("scan_id", "unknown"),
"risk_score": 30,
"risk_trend": "+0",
"analysis": [],
"_skipped": True,
"_reason": "path_C_doc_scan",
}
critic_output = {"verdict": "SKIPPED", "weighted_score": 60.0, "_skipped": True}
_remember_stage("analyst", {"status": "SKIPPED", "reason": "path_C"})
_remember_stage("critic", {"status": "SKIPPED", "reason": "path_C"})
_notify("analyst", "COMPLETE", stages_detail["analyst"])
_notify("critic", "COMPLETE", stages_detail["critic"])
else:
_notify("analyst", "RUNNING", {})
rate_limiter.wait_if_needed("analyst")
analyst_output, analyst_sl = stage_analyst(scout_output, input_type=input_type)
benchmark_context = _build_dim11_benchmark_context(
source=sanitized_stack,
input_type=input_type,
task_plan=task_plan,
code_patterns=sg_code_patterns,
scout_output=scout_output,
extracted_packages=extracted_packages,
)
if benchmark_context:
analyst_output["benchmark_context"] = benchmark_context
analyst_output["observed_cwe_categories"] = benchmark_context["observed_cwe_categories"]
logger.info(
"[DIM11] benchmark context attached before Critic: fixture=%s observed=%d expected=%d",
benchmark_context.get("fixture"),
len(benchmark_context.get("observed_cwe_categories", [])),
len(benchmark_context.get("expected_cwe_categories", [])),
)
analyst_detail = {
"status": "SUCCESS" if not analyst_output.get("_degraded") else "DEGRADED",
"risk_score": analyst_output.get("risk_score", 0),
"duration_ms": analyst_sl.steps[-1].get("duration_ms", 0) if analyst_sl.steps else 0,
}
_remember_stage("analyst", analyst_detail)
_notify("analyst", "COMPLETE", analyst_detail)
from agents.analyst import SKILL_MAP as ANALYST_SKILL_MAP
recorder.stage_enter("analyst", scout_output,
skill_file=ANALYST_SKILL_MAP.get(input_type, "chain_analysis.md"),
input_type=input_type)
recorder.stage_exit("analyst", analyst_detail["status"], analyst_output, analyst_detail["duration_ms"])
_notify("critic", "RUNNING", {})
rate_limiter.wait_if_needed("critic")
critic_output, critic_sl = stage_critic(analyst_output, input_type=input_type)
critic_detail = {
"status": "SUCCESS" if not critic_output.get("_degraded") else "DEGRADED",
"verdict": critic_output.get("verdict", "SKIPPED"),
"score": critic_output.get("weighted_score", 0),
"duration_ms": critic_sl.steps[-1].get("duration_ms", 0) if critic_sl.steps else 0,
}
if critic_output.get("recall_challenge"):
critic_detail["recall_challenge_verdict"] = critic_output["recall_challenge"].get("verdict")
_remember_stage("critic", critic_detail)
_notify("critic", "COMPLETE", critic_detail)
from agents.critic import SKILL_MAP as CRITIC_SKILL_MAP
recorder.stage_enter("critic", analyst_output,
skill_file=CRITIC_SKILL_MAP.get(input_type, "debate_sop.md"),
input_type=input_type)
recorder.stage_exit("critic", critic_detail["status"], critic_output, critic_detail["duration_ms"])
_notify("advisor", "RUNNING", {})
rate_limiter.wait_if_needed("advisor")
advisor_output, advisor_sl = stage_advisor(analyst_output, critic_output, input_type=input_type)
advisor_detail = {
"status": "SUCCESS" if not advisor_output.get("_degraded") else "DEGRADED",
"urgent_count": len(advisor_output.get("actions", {}).get("urgent", [])),
"duration_ms": advisor_sl.steps[-1].get("duration_ms", 0) if advisor_sl.steps else 0,
}
_remember_stage("advisor", advisor_detail)
_notify("advisor", "COMPLETE", advisor_detail)
from agents.advisor import SKILL_MAP as ADVISOR_SKILL_MAP
recorder.stage_enter("advisor", analyst_output,
skill_file=ADVISOR_SKILL_MAP.get(input_type, "action_report.md"),
input_type=input_type)
recorder.stage_exit("advisor", advisor_detail["status"], advisor_output, advisor_detail["duration_ms"])
duration = round(time.time() - pipeline_start, 2)
try:
orch_summary = {}
if orch_ctx is not None:
from agents.orchestrator import finalize_orchestration
orch_ctx.final_confidence = advisor_output.get("confidence", "HIGH")
orch_summary = finalize_orchestration(orch_ctx)
except Exception as exc:
logger.debug("[PIPELINE] finalize_orchestration skipped: %s", exc)
orch_summary = {}
pipeline_meta = {
"pipeline_version": "4.0",
"tech_stack": tech_stack,
"scan_path": scan_path,
"stages_completed": len(completed_stages),
"stages_detail": stages_detail,
"enable_critic": ENABLE_CRITIC,
"critic_verdict": critic_output.get("verdict", "SKIPPED"),
"critic_score": critic_output.get("weighted_score", 0),
"feedback_loops": 0,
"duration_seconds": duration,
"degradation": degradation_status.to_dict(),
"orchestration": orch_summary,
"layer1_agents": [name for name in layer1_results if not name.startswith("_")],
"fusion_after_discovery": bool(task_plan.get("fusion_after_discovery", False)),
"l0_report": l0_report,
"generated_at": datetime.now(timezone.utc).isoformat(),
}
if benchmark_context:
pipeline_meta["benchmark_context"] = benchmark_context
if critic_output.get("recall_challenge"):
pipeline_meta["critic_recall_challenge"] = critic_output["recall_challenge"]
if intel_fusion_result.get("evidence_contract"):
pipeline_meta["intel_fusion_evidence_contract"] = intel_fusion_result["evidence_contract"]
recorder.end_scan("COMPLETE", duration)
report_payload = _build_scan_report_payload(scout_output, advisor_output, layer1_results)
final_output: dict[str, Any] = {
**advisor_output,
**report_payload,
"pipeline_meta": pipeline_meta,
}
if sg_code_patterns:
existing_cps = final_output.get("code_patterns_summary", [])
final_output["code_patterns_summary"] = existing_cps + sg_code_patterns
if critic_output.get("recall_challenge"):
final_output["critic_recall_challenge"] = critic_output["recall_challenge"]
if critic_output.get("needs_rescan"):
final_output["needs_rescan"] = True
return final_output
# ======================================================================
# (comment encoding corrupted)
# ======================================================================
if __name__ == "__main__":
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s [%(levelname)s] %(name)s: %(message)s",
handlers=[logging.StreamHandler(sys.stdout)],
)
if len(sys.argv) > 1:
tech_stack = " ".join(sys.argv[1:])
else:
tech_stack = "Django 4.2, Redis 7.0, PostgreSQL 16"
print(f"\nThreatHunter - Scanning: {tech_stack}\n")
result = run_pipeline(tech_stack)
print("\n=== Pipeline Result ===")
print(json.dumps(result, ensure_ascii=False, indent=2))
|