Spike 22 — Q-017 Trigger-event detection pre-validation CLOSED 2026-05-03 (research-only)

  • Spike ID: Q-and-A-S22 (D4)
  • Group: Group 6 — Deferred-Q pre-validation (Spikes 19-22; 20-22 loose-ordering)
  • Closed: 2026-05-03T11:50 BST
  • Outcome: VALIDATED (research-only; no kill condition by design per plan §7 Task 22)
  • Session: SPIKE-RUNNER session 4 — Batch K (Option B smart-batching: Spike 22 + Spike 21 INLINE same session)
  • Effort: ~⅖d wall-clock (per Batch K override-prompt sharpened estimate; T-file pre-flight ~30min + Tier 2 retrieval ~10min + T-file authoring ~70min + closure cascade in progress)

What this spike delivered

Single artefact: T-file at ~/off-github/library/projects/inherit/T-spike-q017-trigger-event-detection-pre-validation-2026-05-03.md v1.0 (617 lines / 63 KB / 9 body sections / 11 honesty caveats). Pre-validates Q-017 trigger-event detection ahead of Phase 6 InheritWills lock-decision.

14-trigger taxonomy for UK&W IHT Phase-1 SCOPE-DOWN scope (per S8 narrowed corpus ~6-12 statutes):

  1. Marriage (Wills Act 1837 s.18 revocation)
  2. Divorce (Wills Act 1837 s.18A; civil partnership s.18C)
  3. Birth of child (IFPA 1975 s.1; RNRB s.8E)
  4. Birth of grandchild (RNRB lineal-descendant)
  5. Death (testator) — IHTA 1984 s.4 primary IHT trigger; converts PETs to chargeable
  6. Death (named beneficiary) — Wills Act s.33 substitution
  7. Lifetime gift > NRB threshold (IHTA s.3 / s.3A PET; 7y window starts)
  8. PET 7-year-window crossing (IHTA s.7(2) + Sch 1A taper)
  9. Asset acquisition (financial / real-property / company-share / vehicle)
  10. Asset disposal (mirror of 9)
  11. Jurisdiction relocation (testator emigrate/immigrate)
  12. Loss of mental capacity (POA Act 2023 + LPA Regs 2024)
  13. Estate-value band-crossing (NRB / RNRB taper s.8FA £2M-£2.7M)
  14. Pension annuity / drawdown change (HMRC Transformation Roadmap 2027-2028 pension-inclusion)

Detection-mechanism matrix scoring (per §2 of T-file)

TractabilityTriggersSource-data
5 (Phase-1 clean APIs)7, 8, 9a, 9c, 9d, 10a, 10c, 13AISP TrueLayer/Yapily (T59); Companies House Streaming (T52); DVLA VES (T60); internal-state-cron
4 (Phase-1.5 B2B onboarding)9b, 10bHMLR Business Gateway (T60; 4-6wk onboarding)
3 (Phase-1.5 partner-mediated)11, 14Thirdfort jurisdiction relocation (T58); PDP partner-firm-mediated pension (T60)
2 (Year-2+)5 (Phase-1 TUO partial / Year-2+ GP Connect uplift), 10d, 12Tell-Us-Once citizen-mediated (T60); GP Connect (T68); OPG MLPA private-beta (T60)
1 (Year-2+ civil-registry)1, 2, 3, 4, 6GRO (no T-file evidence — gap)

Phase-1 trigger coverage: 9/14 by count (~65%) but ~80%+ by IHT-statutory-weight per S8 6/6 HMRC vectors evidence (triggers 5+7+8+13 carry the entire IHT computation surface; triggers 9-10 modify estate composition).

20-30% gap explicitly enumerated (§4)

  • GRO civil-registry (no T-file evidence): marriage / divorce / birth / death certificates → no public partner-API; ½-1d desk-research recommended
  • HMCTS court-orders (T52 — HMCTS Probate has no public API; family-court adjacent likely no API): decree-absolute / dissolution / financial-orders manual self-report
  • GP Connect Year-2+ (T68): NHS Digital regulated-access partner-firm contract for proactive death-detection
  • OPG MLPA private-beta (T60 §3): 6-9mo ministerial-onboarding-process timeline
  • Multi-jurisdiction-customer fan-out: 8-15 JurisdictionalSubmission records / estate at executor-time (T60-tier2 #80); pre-death equivalent unmeasured

Statutory anchors enumerated (§5)

Wills Act 1837 s.18 / s.18A / s.18C / s.33; IHTA 1984 s.3 / s.3A / s.4 / s.7 / s.8A-8FA / Sch 1A; IFPA 1975 s.1; AEA 1925 intestacy; POA Act 2023 + LPA Regs 2024; PSD2 RTS Article 10 (90-day re-consent HARD operational constraint); Pensions Schemes Act 2021 + HMRC Transformation Roadmap pension-2027-inclusion.

Cost-rows architectural estimates

  • Phase-1 trigger-detection build: ~£8-12K (~15-24 engineer-days = ~3-5wk; 4 detection-adapters + watch-primitive + Catala TriggerEventReview scope + IAS panel)
  • Phase-1.5 HMLR / Thirdfort / PDP extensions: ~£12-25K incremental
  • 5y cumulative additional run-cost: ~£4-7K/yr (per-call AISP refresh + HMLR query volume × testator-active count)

PROV-O annotation pattern (§3.4 of Spike 22 T-file) supports Spike 21 μ.γ event-stream Will-update option (current Will = fold over event-stream of statutorily-significant trigger-events). Trigger-events are first-class auditable artefacts annotated with PROV-O Activity / Entity / Agent semantics + statutory anchor + detection adapter + confidence level.

Cascade artefacts

  • T-file T-spike-q017-trigger-event-detection-pre-validation-2026-05-03.md v1.0 → off-github commit (Batch K batch 1; pending)
  • arch-state v3.32 → v3.33 with §14 placeholder row S22 REPLACED with concrete evidence + Changelog row → docs-strategy commit (Batch K batch 2; pending)
  • Plan v1.11 → v1.12 §0 Spike 22 checkbox ticked + lastmod + CHANGELOG row → docs-strategy commit (same as above; pending)
  • Memory file (this file) saved
  • MEMORY.md index entry pending
  • Active-work-log entry pending CLOSED conversion
  • NO docs-personal commit (no richard-task disposition)

Honesty caveats salient

  1. Research-only — NO prototype built; NO API hit; NO real-world signal tested
  2. Civil-registry (GRO) status NOT directly verified — adjacent T60 evidence only
  3. GP Connect (NHS Digital) status NOT directly verified — T68 framing only
  4. Trigger sensitivity / specificity / latency / cost numbers are architectural estimates at order-of-magnitude
  5. 14-trigger taxonomy covers UK&W IHT Phase-1 SCOPE-DOWN — does NOT enumerate Scotland / CH / multi-jurisdiction

NO modifications

  • A-1..A-142; A-22 SSSOM-canonical pin; A-102 JurisdictionalSubmission stays at-or-above what it accommodates
  • MQ-000 ζ.2 scope; brand-architecture v1.16; master plan v1.10; per-repo BUILD-PLAN v1.4-1.6; refined-prompt v3.7
  • NO new richard-tasks; NO new SKOS classifier scheme; NO new A-21 CI gate (still 22 per A-130); cross-module primitive count UNCHANGED at 27; module count UNCHANGED at 9; risk register UNCHANGED

NEXT (this session)

Spike 21 Batch K closure begins immediately per inline-pair discipline. Closure verifier for Spike 22 will run after Spike 21 closes (single verifier-batch at end of Batch K).

Methodological observations

  • 4th spike of 22-spike suite to close (after Spike 1 + Spike 12 + Spike 18)
  • SECOND research-only spike (after Spike 18)
  • FIRST inline-paired-spike via Batch K Option B smart-batching
  • T-file pre-flight per feedback_check_t_files_first_for_any_inherit_v2_work saved ~50-70% of authoring effort (T59 + T60×2 + T52 + T58 covered ~70% of trigger-detection landscape; civil-registry GAP confirmed)