@@ -0,0 +1,328 @@
+#!/usr/bin/env python3
+"""Build NIST 800-53 v5 eval gold (opencre.org Links) + OSCAL Module A sources.
+
+Gold (CRE answer key)::
+
+  application/tests/fixtures/owasp_mappings/nist_800_53_v5.json
+
+Sources (control statement + guidance from usnistgov OSCAL catalog)::
+
+  scripts/oie_owasp_eval/fixtures/b2_sources/nist_800_53_v5/<SECTION_ID>.txt
+
+Usage::
+
+  PYTHONPATH=. python scripts/oie_owasp_eval/build_nist_800_53_v5_eval.py
+  PYTHONPATH=. python scripts/oie_owasp_eval/build_nist_800_53_v5_eval.py --sample 48
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import sys
+import urllib.parse
+import urllib.request
+from collections import defaultdict
+from pathlib import Path
+from typing import Any, Dict, List, Mapping, Optional, Sequence, Tuple
+
+ROOT = Path(__file__).resolve().parents[2]
+DEFAULT_OUT_GOLD = (
+    ROOT
+    / "application"
+    / "tests"
+    / "fixtures"
+    / "owasp_mappings"
+    / "nist_800_53_v5.json"
+)
+DEFAULT_SOURCES = (
+    ROOT / "scripts" / "oie_owasp_eval" / "fixtures" / "b2_sources" / "nist_800_53_v5"
+)
+OSCAL_URL = (
+    "https://raw.githubusercontent.com/usnistgov/oscal-content/main/"
+    "nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json"
+)
+OSCAL_CACHE = ROOT / "tmp" / "oie_owasp_eval" / "nist_sp80053_rev5_catalog.json"
+BASE_URL = "https://opencre.org"
+USER_AGENT = "OpenCRE-NIST80053-EvalGold/1.0"
+STANDARD_NAME = "NIST 800-53 v5"
+
+# Hub uses AC-2 / IA-2(1); OSCAL labels are AC-02 / IA-02(01).
+_LABEL_RE = re.compile(
+    r"^([A-Z]{1,4})-0*(\d+)(?:\((0*)(\d+)\))?$",
+    re.IGNORECASE,
+)
+
+
+def _get_json(url: str) -> Any:
+    req = urllib.request.Request(url, headers={"User-Agent": USER_AGENT})
+    with urllib.request.urlopen(req, timeout=120) as resp:
+        return json.load(resp)
+
+
+def normalize_control_id(raw: str) -> str:
+    """Map ``AC-02`` / ``IA-02(01)`` / ``ac-2.1`` → hub-style ``AC-2`` / ``IA-2(1)``."""
+    s = (raw or "").strip().upper().replace("_", "-")
+    # OSCAL dotted enhancement id: ia-2.1 / IA-02.01
+    dotted = re.match(r"^([A-Z]{1,4})-0*(\d+)\.0*(\d+)$", s)
+    if dotted:
+        return f"{dotted.group(1)}-{int(dotted.group(2))}({int(dotted.group(3))})"
+    m = _LABEL_RE.match(s)
+    if m:
+        fam = m.group(1).upper()
+        num = int(m.group(2))
+        if m.group(4) is not None:
+            return f"{fam}-{num}({int(m.group(4))})"
+        return f"{fam}-{num}"
+    # Already hub-ish: AC-2(1)
+    m2 = re.match(r"^([A-Z]{1,4})-(\d+)(?:\((\d+)\))?$", s, re.IGNORECASE)
+    if not m2:
+        return s
+    fam, num, enh = m2.group(1).upper(), m2.group(2), m2.group(3)
+    return f"{fam}-{int(num)}" + (f"({int(enh)})" if enh else "")
+
+
+def sid_from_node(node: Mapping[str, Any]) -> Optional[str]:
+    for key in ("sectionID", "section_id"):
+        val = node.get(key)
+        if val:
+            return normalize_control_id(str(val))
+    section = str(node.get("section") or "")
+    m = re.match(r"^([A-Z]{1,4}-\d+(?:\([0-9]+\))?)\b", section, re.I)
+    if m:
+        return normalize_control_id(m.group(1))
+    return None
+
+
+def cre_ids_from_node(node: Mapping[str, Any]) -> List[str]:
+    out: List[str] = []
+    for link in node.get("links") or []:
+        if not isinstance(link, dict):
+            continue
+        doc = link.get("document") or {}
+        if str(doc.get("doctype") or "").upper() != "CRE":
+            continue
+        cid = str(doc.get("id") or "").strip()
+        if cid and cid not in out:
+            out.append(cid)
+    return out
+
+
+def fetch_hub_nodes(*, base_url: str = BASE_URL) -> List[Dict[str, Any]]:
+    rows: List[Dict[str, Any]] = []
+    page = 1
+    while True:
+        url = (
+            f"{base_url.rstrip('/')}/rest/v1/Standard/"
+            f"{urllib.parse.quote(STANDARD_NAME)}?page={page}"
+        )
+        data = _get_json(url)
+        batch = data.get("standards") or []
+        if not batch:
+            break
+        rows.extend(batch)
+        total_pages = int(data.get("total_pages") or page)
+        if page >= total_pages:
+            break
+        page += 1
+    return rows
+
+
+def build_gold_rows(nodes: Sequence[Mapping[str, Any]]) -> List[Dict[str, Any]]:
+    rows: List[Dict[str, Any]] = []
+    seen: set[str] = set()
+    for node in nodes:
+        sid = sid_from_node(node)
+        cres = cre_ids_from_node(node)
+        if not sid or not cres:
+            continue
+        if sid in seen:
+            continue
+        seen.add(sid)
+        rows.append(
+            {
+                "section_id": sid,
+                "section": str(node.get("section") or sid),
+                "hyperlink": str(node.get("hyperlink") or ""),
+                "cre_ids": cres,
+                "gold_provenance": {
+                    "kind": "hub_links",
+                    "method": "opencre_org_standard_api",
+                    "standard": STANDARD_NAME,
+                },
+            }
+        )
+    rows.sort(key=lambda r: r["section_id"])
+    return rows
+
+
+def stratified_sample(
+    rows: Sequence[Dict[str, Any]], n: int
+) -> List[Dict[str, Any]]:
+    """Take up to ``n`` rows, round-robin across control families."""
+    if n <= 0 or n >= len(rows):
+        return list(rows)
+    by_fam: Dict[str, List[Dict[str, Any]]] = defaultdict(list)
+    for row in rows:
+        fam = str(row["section_id"]).split("-", 1)[0]
+        by_fam[fam].append(row)
+    families = sorted(by_fam.keys(), key=lambda f: (-len(by_fam[f]), f))
+    out: List[Dict[str, Any]] = []
+    idx = {f: 0 for f in families}
+    while len(out) < n:
+        progressed = False
+        for fam in families:
+            i = idx[fam]
+            bucket = by_fam[fam]
+            if i >= len(bucket):
+                continue
+            out.append(bucket[i])
+            idx[fam] = i + 1
+            progressed = True
+            if len(out) >= n:
+                break
+        if not progressed:
+            break
+    out.sort(key=lambda r: r["section_id"])
+    for row in out:
+        prov = dict(row.get("gold_provenance") or {})
+        prov["sample"] = True
+        prov["sample_size"] = n
+        row["gold_provenance"] = prov
+    return out
+
+
+def _walk_prose(parts: Sequence[Mapping[str, Any]], chunks: List[str]) -> None:
+    for part in parts or []:
+        if not isinstance(part, dict):
+            continue
+        prose = part.get("prose")
+        if isinstance(prose, str) and prose.strip():
+            chunks.append(prose.strip())
+        nested = part.get("parts")
+        if isinstance(nested, list):
+            _walk_prose(nested, chunks)
+
+
+def load_oscal_index(path: Path) -> Dict[str, Tuple[str, str]]:
+    """Map hub-style section id → (title, statement+guidance prose)."""
+    catalog = json.loads(path.read_text(encoding="utf-8"))["catalog"]
+    index: Dict[str, Tuple[str, str]] = {}
+
+    def add(control: Mapping[str, Any]) -> None:
+        label = None
+        for prop in control.get("props") or []:
+            if isinstance(prop, dict) and prop.get("name") == "label":
+                label = str(prop.get("value") or "")
+                break
+        raw_id = label or str(control.get("id") or "")
+        sid = normalize_control_id(raw_id)
+        title = str(control.get("title") or sid)
+        chunks: List[str] = []
+        # Prefer statement + guidance only (skip assessment noise).
+        for part in control.get("parts") or []:
+            if not isinstance(part, dict):
+                continue
+            name = str(part.get("name") or "")
+            if name not in {"statement", "guidance"}:
+                continue
+            prose = part.get("prose")
+            if isinstance(prose, str) and prose.strip():
+                chunks.append(prose.strip())
+            _walk_prose(part.get("parts") or [], chunks)
+        text = "\n".join(chunks).strip()
+        # Strip OSCAL parameter placeholders for cleaner retrieval.
+        text = re.sub(r"\{\{\s*insert:[^}]+\}\}", "…", text)
+        if sid and text:
+            index[sid] = (title, text)
+        for sub in control.get("controls") or []:
+            if isinstance(sub, dict):
+                add(sub)
+
+    for group in catalog.get("groups") or []:
+        for control in group.get("controls") or []:
+            if isinstance(control, dict):
+                add(control)
+    return index
+
+
+def ensure_oscal(cache: Path = OSCAL_CACHE) -> Path:
+    cache.parent.mkdir(parents=True, exist_ok=True)
+    if cache.is_file() and cache.stat().st_size > 1_000_000:
+        return cache
+    print(f"GET {OSCAL_URL}", flush=True)
+    urllib.request.urlretrieve(OSCAL_URL, cache)
+    return cache
+
+
+def write_sources(
+    rows: Sequence[Mapping[str, Any]],
+    oscal: Mapping[str, Tuple[str, str]],
+    dest: Path,
+) -> Dict[str, int]:
+    dest.mkdir(parents=True, exist_ok=True)
+    written = 0
+    missing = 0
+    for row in rows:
+        sid = str(row["section_id"])
+        title, body = oscal.get(sid, ("", ""))
+        if not body:
+            missing += 1
+            continue
+        section = str(row.get("section") or title or sid)
+        text = (
+            f"Source: {sid}\n"
+            f"Section-ID: {sid}\n"
+            f"Section: {section}\n"
+            f"Standard: {STANDARD_NAME}\n\n"
+            f"{title}\n\n{body}\n"
+        )
+        # Match B2 loader: b2_sources/<stem>/<section_id>.txt
+        (dest / f"{sid}.txt").write_text(text, encoding="utf-8")
+        written += 1
+    return {"written": written, "missing_oscal": missing}
+
+
+def main(argv: Optional[Sequence[str]] = None) -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--out-gold", type=Path, default=DEFAULT_OUT_GOLD)
+    parser.add_argument("--out-sources", type=Path, default=DEFAULT_SOURCES)
+    parser.add_argument(
+        "--sample",
+        type=int,
+        default=0,
+        help="If >0, stratified sample of N controls (default: all hub-linked).",
+    )
+    parser.add_argument("--base-url", default=BASE_URL)
+    parser.add_argument("--skip-sources", action="store_true")
+    args = parser.parse_args(list(argv) if argv is not None else None)
+
+    print(f"fetching {STANDARD_NAME} from {args.base_url} …", flush=True)
+    nodes = fetch_hub_nodes(base_url=args.base_url)
+    gold = build_gold_rows(nodes)
+    print(f"hub nodes={len(nodes)} gold_rows={len(gold)}", flush=True)
+    if args.sample > 0:
+        gold = stratified_sample(gold, args.sample)
+        print(f"sampled to {len(gold)}", flush=True)
+
+    args.out_gold.parent.mkdir(parents=True, exist_ok=True)
+    args.out_gold.write_text(json.dumps(gold, indent=2) + "\n", encoding="utf-8")
+    print(f"wrote {args.out_gold}", flush=True)
+
+    if not args.skip_sources:
+        oscal_path = ensure_oscal()
+        index = load_oscal_index(oscal_path)
+        print(f"oscal controls indexed={len(index)}", flush=True)
+        stats = write_sources(gold, index, args.out_sources)
+        print(json.dumps({"sources_dir": str(args.out_sources), **stats}, indent=2))
+        if stats["missing_oscal"]:
+            print(
+                f"warning: {stats['missing_oscal']} gold ids missing from OSCAL",
+                file=sys.stderr,
+            )
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
@@ -0,0 +1,32 @@
+[
+  {
+    "section_id": "AAI01",
+    "section": "Goal Hijacking and Agent Drift",
+    "hyperlink": "",
+    "cre_ids": ["177-260", "724-770"]
+  },
+  {
+    "section_id": "AAI02",
+    "section": "Tool Misuse and Unsafe Actuation",
+    "hyperlink": "",
+    "cre_ids": ["486-813", "117-371"]
+  },
+  {
+    "section_id": "AAI03",
+    "section": "Multi-Agent Collusion and Privilege Escalation",
+    "hyperlink": "",
+    "cre_ids": ["128-128", "633-428"]
+  },
+  {
+    "section_id": "AAI04",
+    "section": "Memory Poisoning Across Agent Sessions",
+    "hyperlink": "",
+    "cre_ids": ["340-375", "613-287"]
+  },
+  {
+    "section_id": "AAI05",
+    "section": "Unsupervised Cross-System Workflows",
+    "hyperlink": "",
+    "cre_ids": ["132-146", "515-021"]
+  }
+]