feat(ttp): extract intel_lifter provider mappings to YAML data + ATT&CK external_reference enrichment

The four provider→technique tables (AbuseIPDB cat→techniques,
GreyNoise tag→techniques, ThreatFox threat_type→techniques, plus
the Feodo binary-listed signal) used to live as Final[dict] constants
in intel_lifter.py. Two real problems with that:

1. Drift between rules/ttp/R0054.yaml..R0058.yaml (which declare
   the full slate per provider) and the Python dicts (which decide
   which slate-member fires per signal). The v2 audit comment in
   intel_lifter.py documented that they had silently drifted.
2. No ATT&CK provenance on emissions — the loaded STIX bundle has
   rich external_references (canonical attack.mitre.org URLs) that
   never surfaced because the lifter had no path back to them.

Mappings now live as YAML at decnet/ttp/data/intel/{provider}.yaml,
validated at load against the loaded ATT&CK bundle, with each entry
enriched by attack_stix._attack_pattern_by_id to attach the canonical
MITRE URL to every emission.

- decnet/ttp/data/intel_loader.py: pydantic-validated schema +
  ProviderMapping/Signal/TechniqueEmission frozen dataclasses +
  load_provider_mapping(provider) lru-cached.
- Per-technique high_score_threshold inlined into YAML
  (collapses the separate _ABUSEIPDB_HIGH_SCORE_GATED dict).
- external_reference field follows the STIX 2.1 external-reference
  shape (source_name + url + optional external_id) so the future
  STIX/MISP exporter is a direct translation.
- intel_lifter.py: dicts deleted, decision functions read from
  ProviderMapping accessors. Decision-flow constants (T1071/T1595
  bare-classification fallbacks in _greynoise_decisions) stay in
  code — they're not table rows.
- Each emit slot's evidence_extra now carries mitre_url for any
  technique resolved in the bundle (every one in practice).
- tests/ttp/test_intel_mappings.py: snapshot equivalence vs the
  legacy dicts, high-score gate behavior, every-signal-has-an-
  external-reference, every-emission-has-a-mitre-url, negative
  paths (unknown technique_id raises AttackBundleError, mismatched
  provider field rejected, dir listing matches expected providers).

The YAML schema + mitre_url enrichment lays groundwork for the
future STIX exporter; this commit does NOT build that exporter.

decnet/ttp/data/__init__.py

@@ -0,0 +1,6 @@
+"""Data files used at runtime by the TTP layer.
+
+See ``decnet/ttp/data/intel/`` for provider-signal → ATT&CK technique
+mappings consumed by :mod:`decnet.ttp.impl.intel_lifter` via
+:mod:`decnet.ttp.data.intel_loader`.
+"""

decnet/ttp/data/intel/__init__.py

@@ -0,0 +1,8 @@
+"""Per-provider intel-signal → ATT&CK technique mapping data.
+
+One YAML file per intel provider (abuseipdb / greynoise / feodo /
+threatfox), structured per the schema in
+:mod:`decnet.ttp.data.intel_loader`. Each entry carries a STIX-shaped
+``external_reference`` so the future STIX/MISP exporter can emit
+relationship objects without a second mapping pass.
+"""

decnet/ttp/data/intel/abuseipdb.yaml

@@ -0,0 +1,125 @@
+# AbuseIPDB category → ATT&CK technique mapping.
+#
+# Mirrors what _ABUSEIPDB_CATEGORY_TO_TECHNIQUES + _ABUSEIPDB_HIGH_SCORE_GATED
+# used to encode in decnet/ttp/impl/intel_lifter.py before the data
+# extraction. Source-of-truth column for which categories produce
+# which ATT&CK tags, paired with rules/ttp/R0054.yaml which declares
+# the full slate the predicate can emit.
+#
+# Cat 4 (DDoS), 10 (Web Spam), 12 (Blog Spam) are intentionally
+# unmapped — design doc TTP_TAGGING.md §A.10: DDoS-without-protocol
+# is too muddy for v0; CMS spam has no clean ATT&CK fit at the IP
+# layer. Keep the explanatory comments here so the next quarterly
+# drift check (development/DEBT.md DEBT-048) can diff cheaply.
+provider: abuseipdb
+mapping_version: "2"
+attack_release: ">=15.1"
+signals:
+  - id: cat_5
+    label: "FTP Brute-Force"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#5"
+    techniques:
+      - technique_id: T1110
+  - id: cat_7
+    label: "Phishing"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#7"
+    techniques:
+      - technique_id: T1566
+  - id: cat_9
+    label: "Open Proxy"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#9"
+    techniques:
+      - technique_id: T1090
+  - id: cat_11
+    label: "Email Spam"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#11"
+    techniques:
+      - technique_id: T1496
+      - technique_id: T1566
+        high_score_threshold: 80
+  - id: cat_13
+    label: "VPN IP"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#13"
+    techniques:
+      - technique_id: T1090
+  - id: cat_14
+    label: "Port Scan"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#14"
+    techniques:
+      - technique_id: T1046
+      - technique_id: T1595
+  - id: cat_15
+    label: "Hacking"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#15"
+    techniques:
+      - technique_id: T1190
+  - id: cat_16
+    label: "SQL Injection"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#16"
+    techniques:
+      - technique_id: T1190
+  - id: cat_17
+    label: "Spoofing"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#17"
+    techniques:
+      - technique_id: T1566
+  - id: cat_18
+    label: "Brute-Force"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#18"
+    techniques:
+      - technique_id: T1110
+  - id: cat_19
+    label: "Bad Web Bot"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#19"
+    techniques:
+      - technique_id: T1595
+  - id: cat_20
+    label: "Exploited Host"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#20"
+    techniques:
+      - technique_id: T1078
+  - id: cat_21
+    label: "Web App Attack"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#21"
+    techniques:
+      - technique_id: T1190
+  - id: cat_22
+    label: "SSH"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#22"
+    techniques:
+      - technique_id: T1110
+  - id: cat_23
+    label: "IoT Targeted"
+    external_reference:
+      source_name: abuseipdb
+      url: "https://www.abuseipdb.com/categories#23"
+    techniques:
+      - technique_id: T1190

decnet/ttp/data/intel/feodo.yaml

@@ -0,0 +1,20 @@
+# Feodo Tracker → ATT&CK technique mapping.
+#
+# Feodo Tracker is a binary listed/not-listed feed; there are no
+# per-signal subtypes to enumerate. Both T1071 (Application Layer
+# Protocol) and T1588 (Obtain Capabilities) fire whenever an attacker
+# IP is on the Feodo blocklist. Keeping this as a single ``feodo_listed``
+# signal preserves the structured-mapping shape for the future
+# STIX/MISP exporter without inventing fake categories.
+provider: feodo
+mapping_version: "1"
+attack_release: ">=15.1"
+signals:
+  - id: feodo_listed
+    label: "Listed on Feodo Tracker"
+    external_reference:
+      source_name: feodo
+      url: "https://feodotracker.abuse.ch/about/"
+    techniques:
+      - technique_id: T1071
+      - technique_id: T1588

decnet/ttp/data/intel/greynoise.yaml

@@ -0,0 +1,74 @@
+# GreyNoise tag → ATT&CK technique mapping.
+#
+# Mirrors what _GREYNOISE_TAG_TO_TECHNIQUES used to encode in
+# decnet/ttp/impl/intel_lifter.py. Note: GreyNoise's Community
+# endpoint does not return tags; these fire only when operators wire
+# a non-Community provider (Visualizer / Enterprise / RIOT). Kept
+# canonical here so the upgrade path is a column populate, not a
+# code change. Decision-flow constants for bare ``classification ==
+# "scanner"`` (T1595) and bare ``classification == "malicious"``
+# (T1071 at 0.5×) stay in code — they're not table rows.
+provider: greynoise
+mapping_version: "1"
+attack_release: ">=15.1"
+signals:
+  - id: tor_exit_node
+    label: "Tor exit node"
+    external_reference:
+      source_name: greynoise
+      url: "https://docs.greynoise.io/docs/understanding-greynoise-tags"
+      external_id: tor_exit_node
+    techniques:
+      - technique_id: T1090
+  - id: ssh_bruteforcer
+    label: "SSH brute-forcer"
+    external_reference:
+      source_name: greynoise
+      url: "https://docs.greynoise.io/docs/understanding-greynoise-tags"
+      external_id: ssh_bruteforcer
+    techniques:
+      - technique_id: T1110
+  - id: web_crawler
+    label: "Web crawler"
+    external_reference:
+      source_name: greynoise
+      url: "https://docs.greynoise.io/docs/understanding-greynoise-tags"
+      external_id: web_crawler
+    techniques:
+      - technique_id: T1595
+  - id: cobalt_strike
+    label: "Cobalt Strike"
+    external_reference:
+      source_name: greynoise
+      url: "https://docs.greynoise.io/docs/understanding-greynoise-tags"
+      external_id: cobalt_strike
+    techniques:
+      - technique_id: T1071
+      - technique_id: T1588
+  - id: metasploit
+    label: "Metasploit"
+    external_reference:
+      source_name: greynoise
+      url: "https://docs.greynoise.io/docs/understanding-greynoise-tags"
+      external_id: metasploit
+    techniques:
+      - technique_id: T1071
+      - technique_id: T1588
+  - id: sliver
+    label: "Sliver"
+    external_reference:
+      source_name: greynoise
+      url: "https://docs.greynoise.io/docs/understanding-greynoise-tags"
+      external_id: sliver
+    techniques:
+      - technique_id: T1071
+      - technique_id: T1588
+  - id: havoc
+    label: "Havoc"
+    external_reference:
+      source_name: greynoise
+      url: "https://docs.greynoise.io/docs/understanding-greynoise-tags"
+      external_id: havoc
+    techniques:
+      - technique_id: T1071
+      - technique_id: T1588

decnet/ttp/data/intel/threatfox.yaml

@@ -0,0 +1,45 @@
+# ThreatFox threat_type → ATT&CK technique mapping.
+#
+# Mirrors _THREATFOX_THREAT_TYPE_TO_TECHNIQUES from
+# decnet/ttp/impl/intel_lifter.py. ThreatFox's canonical taxonomy is
+# the ``threat_type`` field (NOT ``ioc_type`` — that was the v1
+# ship-time bug). ``ioc_type`` is the indicator format (url, domain,
+# md5_hash, …) and carries no ATT&CK signal.
+provider: threatfox
+mapping_version: "1"
+attack_release: ">=15.1"
+signals:
+  - id: botnet_cc
+    label: "Botnet C2"
+    external_reference:
+      source_name: threatfox
+      url: "https://threatfox.abuse.ch/faq/"
+      external_id: botnet_cc
+    techniques:
+      - technique_id: T1071
+      - technique_id: T1588
+  - id: payload_delivery
+    label: "Payload delivery"
+    external_reference:
+      source_name: threatfox
+      url: "https://threatfox.abuse.ch/faq/"
+      external_id: payload_delivery
+    techniques:
+      - technique_id: T1105
+      - technique_id: T1588
+  - id: payload
+    label: "Payload"
+    external_reference:
+      source_name: threatfox
+      url: "https://threatfox.abuse.ch/faq/"
+      external_id: payload
+    techniques:
+      - technique_id: T1588
+  - id: cc_skimming
+    label: "Credit-card skimming"
+    external_reference:
+      source_name: threatfox
+      url: "https://threatfox.abuse.ch/faq/"
+      external_id: cc_skimming
+    techniques:
+      - technique_id: T1056

decnet/ttp/data/intel_loader.py

@@ -0,0 +1,229 @@
+"""YAML-backed loader for intel-provider → ATT&CK technique mappings.
+
+Replaces the ``_*_TO_TECHNIQUES`` ``Final[dict]`` tables that used to
+live in :mod:`decnet.ttp.impl.intel_lifter`. Source-of-truth files
+live under :mod:`decnet.ttp.data.intel` (one YAML per provider) and
+are validated against the loaded ATT&CK STIX bundle at load time:
+
+* every ``technique_id`` in every signal must resolve in
+  :func:`decnet.ttp.attack_stix.technique_exists`
+* every entry is enriched with the canonical MITRE
+  ``external_reference`` (source_name=``mitre-attack``, url) so the
+  future STIX/MISP exporter can emit fully-resolved relationship
+  objects without a second mapping pass
+
+Design constraint: this module is the only place provider-mapping
+schema knowledge lives. ``intel_lifter`` reads :class:`ProviderMapping`
+accessors and never touches the dicts directly.
+"""
+from __future__ import annotations
+
+from dataclasses import dataclass
+from functools import lru_cache
+from pathlib import Path
+from typing import Any
+
+import yaml
+from pydantic import BaseModel, ConfigDict, Field
+
+from decnet.ttp import attack_stix
+
+_DATA_DIR: Path = Path(__file__).parent / "intel"
+
+
+# ─── YAML schema (pydantic v2) ─────────────────────────────────────
+
+
+class ExternalReference(BaseModel):
+    """STIX 2.1 ``external-reference`` shape — kept faithful so the
+    future STIX exporter is a direct translation."""
+
+    model_config = ConfigDict(frozen=True)
+
+    source_name: str
+    url: str
+    external_id: str | None = None
+
+
+class TechniqueEntry(BaseModel):
+    model_config = ConfigDict(frozen=True)
+
+    technique_id: str
+    # Per-technique gate: emission only fires when an upstream
+    # confidence score (e.g. AbuseIPDB ``abuseConfidenceScore``)
+    # meets or exceeds this floor. None = always fire.
+    high_score_threshold: int | None = None
+
+
+class SignalEntry(BaseModel):
+    model_config = ConfigDict(frozen=True)
+
+    id: str
+    label: str
+    external_reference: ExternalReference
+    techniques: tuple[TechniqueEntry, ...]
+    confidence_multiplier: float = 1.0
+
+
+class ProviderMappingFile(BaseModel):
+    model_config = ConfigDict(frozen=True)
+
+    provider: str
+    mapping_version: str
+    attack_release: str = Field(
+        description="Minimum ATT&CK release this mapping is known-correct against."
+    )
+    signals: tuple[SignalEntry, ...]
+
+
+# ─── Runtime accessor objects ──────────────────────────────────────
+
+
+@dataclass(frozen=True)
+class TechniqueEmission:
+    """A single emit slot for a (signal, technique) pair, enriched with the canonical MITRE URL."""
+
+    technique_id: str
+    high_score_threshold: int | None
+    mitre_url: str | None
+
+
+@dataclass(frozen=True)
+class Signal:
+    id: str
+    label: str
+    external_reference: ExternalReference
+    emissions: tuple[TechniqueEmission, ...]
+    confidence_multiplier: float
+
+    def technique_ids(self) -> frozenset[str]:
+        return frozenset(e.technique_id for e in self.emissions)
+
+
+@dataclass(frozen=True)
+class ProviderMapping:
+    provider: str
+    mapping_version: str
+    signals: tuple[Signal, ...]
+    _by_id: dict[str, Signal]
+
+    def get(self, signal_id: str) -> Signal | None:
+        return self._by_id.get(signal_id)
+
+    def techniques_for_signal(
+        self, signal_id: str, *, score: float | None = None
+    ) -> frozenset[TechniqueEmission]:
+        """Emissions a given signal produces, filtered by ``score``-vs-threshold gate.
+
+        ``score`` is the upstream confidence (e.g. AbuseIPDB
+        ``abuseConfidenceScore`` 0-100). If a technique has a
+        ``high_score_threshold`` and ``score`` is below it (or
+        unknown), that technique is filtered out. Mirrors the legacy
+        ``_ABUSEIPDB_HIGH_SCORE_GATED`` semantics.
+        """
+        sig = self._by_id.get(signal_id)
+        if sig is None:
+            return frozenset()
+        out: set[TechniqueEmission] = set()
+        for emission in sig.emissions:
+            if emission.high_score_threshold is not None:
+                if score is None or score < emission.high_score_threshold:
+                    continue
+            out.add(emission)
+        return frozenset(out)
+
+    def all_technique_ids(self) -> frozenset[str]:
+        return frozenset(
+            e.technique_id for sig in self.signals for e in sig.emissions
+        )
+
+    def signal_ids(self) -> frozenset[str]:
+        return frozenset(self._by_id.keys())
+
+
+# ─── Loader ────────────────────────────────────────────────────────
+
+
+def _mitre_url_for(technique_id: str) -> str | None:
+    obj = attack_stix._attack_pattern_by_id(technique_id)
+    if obj is None:
+        return None
+    for ref in obj.get("external_references", []):
+        if ref.get("source_name") == "mitre-attack":
+            return ref.get("url")
+    return None
+
+
+def _data_path(provider: str) -> Path:
+    return _DATA_DIR / f"{provider}.yaml"
+
+
+@lru_cache(maxsize=8)
+def load_provider_mapping(provider: str) -> ProviderMapping:
+    """Load + validate + enrich a provider's mapping YAML. Cached process-wide."""
+    path = _data_path(provider)
+    if not path.is_file():
+        raise FileNotFoundError(
+            f"intel mapping for provider {provider!r} not found at {path}"
+        )
+    raw: Any = yaml.safe_load(path.read_text(encoding="utf-8"))
+    parsed = ProviderMappingFile.model_validate(raw)
+    if parsed.provider != provider:
+        raise ValueError(
+            f"{path}: provider field {parsed.provider!r} does not match "
+            f"filename {provider!r}"
+        )
+
+    # Validate every technique resolves in the loaded ATT&CK bundle.
+    all_ids = sorted(
+        {t.technique_id for s in parsed.signals for t in s.techniques}
+    )
+    attack_stix.assert_known_technique_ids(
+        all_ids, source=f"decnet/ttp/data/intel/{provider}.yaml"
+    )
+
+    signals: list[Signal] = []
+    for s in parsed.signals:
+        emissions = tuple(
+            TechniqueEmission(
+                technique_id=t.technique_id,
+                high_score_threshold=t.high_score_threshold,
+                mitre_url=_mitre_url_for(t.technique_id),
+            )
+            for t in s.techniques
+        )
+        signals.append(
+            Signal(
+                id=s.id,
+                label=s.label,
+                external_reference=s.external_reference,
+                emissions=emissions,
+                confidence_multiplier=s.confidence_multiplier,
+            )
+        )
+    by_id = {s.id: s for s in signals}
+    if len(by_id) != len(signals):
+        dupes = [s.id for s in signals if list(by_id).count(s.id) > 1]
+        raise ValueError(f"{path}: duplicate signal ids: {dupes}")
+
+    return ProviderMapping(
+        provider=parsed.provider,
+        mapping_version=parsed.mapping_version,
+        signals=tuple(signals),
+        _by_id=by_id,
+    )
+
+
+def clear_cache() -> None:
+    """Drop cached :class:`ProviderMapping` instances. Test-only knob."""
+    load_provider_mapping.cache_clear()
+
+
+__all__ = [
+    "ExternalReference",
+    "ProviderMapping",
+    "Signal",
+    "TechniqueEmission",
+    "clear_cache",
+    "load_provider_mapping",
+]

decnet/ttp/impl/intel_lifter.py

@@ -17,9 +17,15 @@ gate emission, not provider count).
 from __future__ import annotations
 
 from collections.abc import Callable
+from functools import lru_cache
 from typing import Any, Final
 
 from decnet.ttp.base import TaggerEvent, TolerantTagger
+from decnet.ttp.data.intel_loader import (
+    ProviderMapping,
+    TechniqueEmission,
+    load_provider_mapping,
+)
 from decnet.ttp.impl._emit import emit_tags
 from decnet.ttp.impl._rule_index import RuleIndex
 from decnet.ttp.impl._state import apply_ceiling, is_active
@@ -28,71 +34,39 @@ from decnet.ttp.store.base import RuleStore
 from decnet.web.db.models.ttp import TTPTag, compute_tag_uuid
 
 
-# AbuseIPDB category → set of technique_ids that fire on it. Derived
-# from TTP_TAGGING.md Appendix A.10 (post 2026-05-02 ship-time audit).
-# Category code names are AbuseIPDB's canonical taxonomy at
-# https://www.abuseipdb.com/categories — kept verbatim in the comment so
-# the next quarterly drift check (development/DEBT.md DEBT-048) can
-# diff cheaply. Cat 4 (DDoS Attack) and 10 (Web Spam) and 12 (Blog
-# Spam) are intentionally unmapped — design doc §A.10 marks
-# DDoS-without-protocol as too muddy for v0, and CMS spam has no clean
-# ATT&CK fit at the IP layer.
-_ABUSEIPDB_CATEGORY_TO_TECHNIQUES: Final[dict[int, frozenset[str]]] = {
-    5: frozenset({"T1110"}),                  # FTP Brute-Force
-    7: frozenset({"T1566"}),                  # Phishing
-    9: frozenset({"T1090"}),                  # Open Proxy
-    11: frozenset({"T1496", "T1566"}),        # Email Spam (T1566 high-score only)
-    13: frozenset({"T1090"}),                 # VPN IP
-    14: frozenset({"T1046", "T1595"}),        # Port Scan
-    15: frozenset({"T1190"}),                 # Hacking
-    16: frozenset({"T1190"}),                 # SQL Injection
-    17: frozenset({"T1566"}),                 # Spoofing (email-sender)
-    18: frozenset({"T1110"}),                 # Brute-Force
-    19: frozenset({"T1595"}),                 # Bad Web Bot
-    20: frozenset({"T1078"}),                 # Exploited Host
-    21: frozenset({"T1190"}),                 # Web App Attack
-    22: frozenset({"T1110"}),                 # SSH
-    23: frozenset({"T1190"}),                 # IoT Targeted
-}
-
-# Categories where a technique only fires above a confidence-score
-# threshold (per A.10: "11 — Email Spam (high score, ≥80) → T1566").
-_ABUSEIPDB_HIGH_SCORE_GATED: Final[dict[int, dict[str, int]]] = {
-    11: {"T1566": 80},
-}
+# Provider→technique mappings live as YAML under
+# decnet/ttp/data/intel/{provider}.yaml — see
+# decnet.ttp.data.intel_loader for the schema and validation. Lazy
+# accessors below mean module import does not trigger an ATT&CK
+# bundle load (the loader validates every technique resolves there).
 
 
-# GreyNoise tag → set of technique_ids the tag warrants. Note: the
-# Community endpoint does not return tags today — these fire only when
-# operators wire a non-Community provider that does. Kept canonical so
-# the upgrade path is just a column populate, not a code change.
-_GREYNOISE_TAG_TO_TECHNIQUES: Final[dict[str, frozenset[str]]] = {
-    "tor_exit_node": frozenset({"T1090"}),
-    "ssh_bruteforcer": frozenset({"T1110"}),
-    "web_crawler": frozenset({"T1595"}),
-    "cobalt_strike": frozenset({"T1071", "T1588"}),
-    "metasploit": frozenset({"T1071", "T1588"}),
-    "sliver": frozenset({"T1071", "T1588"}),
-    "havoc": frozenset({"T1071", "T1588"}),
-}
+@lru_cache(maxsize=4)
+def _mapping(provider: str) -> ProviderMapping:
+    return load_provider_mapping(provider)
+
 
 # Confidence multiplier when GreyNoise reports ``classification ==
 # "malicious"`` without a specific tag we recognise. The bare
 # classification is real signal but weaker than a tag — half-confidence
-# keeps the floor honest.
+# keeps the floor honest. Decision-flow constant, not a table row.
 _GREYNOISE_MALICIOUS_BARE_MULT: Final[float] = 0.5
 
-# ThreatFox THREAT TYPE (NOT ioc_type — that was the v1 ship-time bug)
-# → set of technique_ids. Per ThreatFox's API the canonical taxonomy
-# field is ``threat_type`` ∈ {botnet_cc, payload_delivery, payload,
-# cc_skimming}; ``ioc_type`` is the indicator format (url, domain,
-# md5_hash, …) and carries no ATT&CK signal.
-_THREATFOX_THREAT_TYPE_TO_TECHNIQUES: Final[dict[str, frozenset[str]]] = {
-    "botnet_cc": frozenset({"T1071", "T1588"}),
-    "payload_delivery": frozenset({"T1105", "T1588"}),
-    "payload": frozenset({"T1588"}),
-    "cc_skimming": frozenset({"T1056"}),
-}
+
+def _emission_url_extras(
+    emissions: dict[str, TechniqueEmission],
+) -> dict[str, dict[str, str]]:
+    """Map technique_id → {"mitre_url": "<url>"} for every emission that has one.
+
+    Lets the per-decision-function evidence_extra builders attach the
+    canonical MITRE URL to each emit slot without re-resolving against
+    the loaded ATT&CK bundle.
+    """
+    return {
+        tid: {"mitre_url": e.mitre_url}
+        for tid, e in emissions.items()
+        if e.mitre_url
+    }
 
 
 # Predicate signature: returns either a list of (technique_id_filter,
@@ -114,14 +88,15 @@ def _abuseipdb_decisions(
     categories: list[int] = [c for c in categories_raw if isinstance(c, int)]
     if not categories:
         return []
-    # Resolve technique set across all categories present.
+    mapping = _mapping("abuseipdb")
+    # Resolve technique set across all categories present, applying
+    # any per-technique high-score gate (see TechniqueEmission).
     triggered: dict[str, list[int]] = {}
+    emissions_by_tech: dict[str, TechniqueEmission] = {}
     for cat in categories:
-        for tech in _ABUSEIPDB_CATEGORY_TO_TECHNIQUES.get(cat, frozenset()):
-            gate = _ABUSEIPDB_HIGH_SCORE_GATED.get(cat, {}).get(tech)
-            if gate is not None and score < gate:
-                continue
-            triggered.setdefault(tech, []).append(cat)
+        for emission in mapping.techniques_for_signal(f"cat_{cat}", score=float(score)):
+            triggered.setdefault(emission.technique_id, []).append(cat)
+            emissions_by_tech.setdefault(emission.technique_id, emission)
     if not triggered:
         return []
     multiplier = float(score) / 100.0
@@ -129,6 +104,10 @@ def _abuseipdb_decisions(
         (tech, multiplier, {
             "abuseipdb_categories": cats,
             "abuse_confidence_score": int(score),
+            **(
+                {"mitre_url": emissions_by_tech[tech].mitre_url}
+                if emissions_by_tech[tech].mitre_url else {}
+            ),
         })
         for tech, cats in triggered.items()
     ]
@@ -152,20 +131,26 @@ def _greynoise_decisions(
     """
     classification = payload.get("greynoise_classification")
     tags_raw = payload.get("greynoise_tags") or []
+    mapping = _mapping("greynoise")
     # Per-technique evidence accumulator — maps technique_id to the
     # signals that triggered it AND the multiplier to apply (max wins
     # if multiple lanes hit the same technique).
     triggered: dict[str, tuple[float, list[str]]] = {}
+    emissions_by_tech: dict[str, TechniqueEmission] = {}
 
-    def _bump(tech: str, mult: float, signal: str) -> None:
+    def _bump(
+        tech: str, mult: float, signal: str, emission: TechniqueEmission | None = None,
+    ) -> None:
         existing = triggered.get(tech)
         if existing is None:
             triggered[tech] = (mult, [signal])
-            return
-        old_mult, signals = existing
-        signals.append(signal)
-        if mult > old_mult:
-            triggered[tech] = (mult, signals)
+        else:
+            old_mult, signals = existing
+            signals.append(signal)
+            if mult > old_mult:
+                triggered[tech] = (mult, signals)
+        if emission is not None:
+            emissions_by_tech.setdefault(tech, emission)
 
     if classification == "scanner":
         _bump("T1595", 1.0, "scanner")
@@ -173,8 +158,8 @@ def _greynoise_decisions(
         for tag in tags_raw:
             if not isinstance(tag, str):
                 continue
-            for tech in _GREYNOISE_TAG_TO_TECHNIQUES.get(tag, frozenset()):
-                _bump(tech, 1.0, tag)
+            for emission in mapping.techniques_for_signal(tag):
+                _bump(emission.technique_id, 1.0, tag, emission)
     if classification == "malicious" and "T1071" not in triggered:
         _bump("T1071", _GREYNOISE_MALICIOUS_BARE_MULT, "malicious")
     if not triggered:
@@ -183,6 +168,11 @@ def _greynoise_decisions(
         (tech, mult, {
             "greynoise_classification": classification,
             "greynoise_tags": signals,
+            **(
+                {"mitre_url": emissions_by_tech[tech].mitre_url}
+                if tech in emissions_by_tech and emissions_by_tech[tech].mitre_url
+                else {}
+            ),
         })
         for tech, (mult, signals) in triggered.items()
     ]
@@ -197,14 +187,17 @@ def _feodo_decisions(
         payload.get("feodo_malware_family")
         or payload.get("malware_family")
     )
-    extra: dict[str, Any] = {"feodo_listed": True}
+    base_extra: dict[str, Any] = {"feodo_listed": True}
     if isinstance(family, str) and family:
-        extra["malware_family"] = family
-    # Both T1071 and T1588 emits fire from a Feodo hit.
-    return [
-        ("T1071", 1.0, extra),
-        ("T1588", 1.0, extra),
-    ]
+        base_extra["malware_family"] = family
+    mapping = _mapping("feodo")
+    out: EmitDecision = []
+    for emission in mapping.techniques_for_signal("feodo_listed"):
+        extra = dict(base_extra)
+        if emission.mitre_url:
+            extra["mitre_url"] = emission.mitre_url
+        out.append((emission.technique_id, 1.0, extra))
+    return out
 
 
 def _threatfox_decisions(
@@ -229,10 +222,13 @@ def _threatfox_decisions(
     elif isinstance(threat_types_raw, str) and threat_types_raw:
         threat_types = [threat_types_raw]
 
+    mapping = _mapping("threatfox")
     triggered: dict[str, list[str]] = {}
+    emissions_by_tech: dict[str, TechniqueEmission] = {}
     for tt in threat_types:
-        for tech in _THREATFOX_THREAT_TYPE_TO_TECHNIQUES.get(tt, frozenset()):
-            triggered.setdefault(tech, []).append(tt)
+        for emission in mapping.techniques_for_signal(tt):
+            triggered.setdefault(emission.technique_id, []).append(tt)
+            emissions_by_tech.setdefault(emission.technique_id, emission)
     if not triggered:
         return []
 
@@ -256,6 +252,10 @@ def _threatfox_decisions(
             "threat_types": signals,
             **({"malware_families": families} if families else {}),
             **({"ioc_types": ioc_types} if ioc_types else {}),
+            **(
+                {"mitre_url": emissions_by_tech[tech].mitre_url}
+                if emissions_by_tech[tech].mitre_url else {}
+            ),
         })
         for tech, signals in triggered.items()
     ]
@@ -376,25 +376,22 @@ def _emit_filtered(
 
 
 def all_emitted_technique_ids() -> frozenset[str]:
-    """Every technique ID this lifter could emit, drawn from all four provider tables.
+    """Every technique ID this lifter could emit, drawn from the four provider mapping YAMLs plus decision-flow constants.
 
-    Used by :func:`validate_against_attack_bundle` (and
-    :mod:`tests.ttp.test_attack_catalog`-adjacent tests) to assert that
-    every provider-driven emission resolves in the loaded ATT&CK STIX
+    Used by :func:`validate_against_attack_bundle` to assert every
+    provider-driven emission resolves in the loaded ATT&CK STIX
     bundle. Includes the bare-classification emissions in
-    ``_greynoise_decisions`` and the unconditional emissions in
-    ``_feodo_decisions`` — those don't appear in the lookup tables
-    above because they're decision-flow constants, not table entries.
+    ``_greynoise_decisions`` (T1595 for ``classification == "scanner"``,
+    T1071 for bare ``"malicious"``) — those are decision-flow
+    constants, not YAML rows. The loader itself already validates
+    YAML-sourced IDs at load; this fold-in covers the in-code
+    constants too.
     """
     ids: set[str] = set()
-    for techs in _ABUSEIPDB_CATEGORY_TO_TECHNIQUES.values():
-        ids.update(techs)
-    for techs in _GREYNOISE_TAG_TO_TECHNIQUES.values():
-        ids.update(techs)
-    for techs in _THREATFOX_THREAT_TYPE_TO_TECHNIQUES.values():
-        ids.update(techs)
-    # Decision-flow constants (see _greynoise_decisions, _feodo_decisions).
-    ids.update({"T1071", "T1595", "T1588"})
+    for provider in ("abuseipdb", "greynoise", "feodo", "threatfox"):
+        ids.update(_mapping(provider).all_technique_ids())
+    # Decision-flow constants (see _greynoise_decisions).
+    ids.update({"T1071", "T1595"})
     return frozenset(ids)