TT Claude Memory

feedback_v3_15_release_v6_6_substrate_consultation_clean_break_2026_05_17

Properties1
descriptionRefined-end-of-turn-directive v3.15 adds step 6h V6.6 SUBSTRATE CONSULTATION as a mandatory research-sweep sub-bullet. v6.6 (~/openinherit/code-standard/) is RESEARCH-SOURCE-ONLY for v2 — read v6.6 to understand legal substrate; DISTIL into v2-native vocabulary; NEVER cite v6.6 paths / $id URLs / module names in v2 outputs (cascade-Q files / scorecards / sub_clarifications_locked / arch-state amendments / spike-references / commit messages). Reason — v6.6 will be deleted post-v2 release per brand-architecture build-order; v2 must be SELF-CONTAINED clean-room for acquirer-DD + standards-engagement positioning. RC7 added to research-completeness self-score in step 6i. Rich-directive 2026-05-17 BST after Q-027 retrospective identified ~2700L of pre-authored multi-jurisdiction substrate (Switzerland + Scotland + UK&W + Hindu + Islamic + Jewish + jurisdiction-profiles) not systematically consulted.

4 min read

v3.15 release — v6.6 substrate consultation + clean-break discipline (2026-05-17)

The rule

For any v2 cascade-Q touching schema / classifier / jurisdiction / mechanism / regulatory references / faith-tradition / form-requirements / professional-terminology:

  1. READ relevant v6.6 substrate files from ~/openinherit/code-standard/ during step 6 RESEARCH SWEEP — specifically reference-data/jurisdiction-profiles.json + v3/extensions/<jurisdiction>/ + reference-data/*.json + extensions-registry.json.
  2. DISTIL the legal substrate into v2-native universal-classifier vocabulary — capture the underlying legal mechanism (e.g. Swiss Pflichtteil 2023 reform; Scotland legal-rights jus relictae/jus relicti/legitim; Hindu HSA 1956 coparcenary; Muslim faraid; etc.) not the v6.6 schema implementation of it.
  3. CITE source law directly in v2 outputs — ZGB Art 470+ / Succession (Scotland) Act 1964 / T&S(S)A 2024 / Wills Act 1837 / Hindu Succession Act 1956 / etc. Source law citations are STABLE across v6.6 deprecation.
  4. NEVER cite v6.6 paths / file names / $id URLs (openinherit.org/v3/...) / inheritVersion strings (>=3.0.0 <4.0.0) / v6.6 module names in v2 outputs. The v2 codebase MUST stand alone post-v6.6-deletion.

Why

  • v6.6 will be deleted post-v2 release per brand-architecture.md v1.16+ build-order discipline. If v2 outputs reference v6.6 paths, those references break at deletion.
  • Acquirer-DD strength — a v2 standard that referenced v6.6 paths would look like a half-finished migration. A clean-room v2 demonstrably stands alone.
  • Standards-engagement positioning per A-141 (OASIS LegalDocML TC / OIDF eKYC&IDA WG / OIDF DADE CG / W3C VC) — external standards bodies cannot evaluate or adopt a v2 that has internal v6.6 path dependencies.
  • Research efficiency — v6.6 has ~2700L of pre-authored multi-jurisdiction substrate. Without this discipline, v2 Qs re-derive substrate v6.6 already encodes (waste). With this discipline, v2 Qs reach jurisdiction-symmetric substrate quickly while staying clean-room.

How to apply

  • During step 6 RESEARCH SWEEP, when the Q’s topic-keyword matches a v6.6 substrate spine (jurisdiction / classifier / mechanism / regulatory / faith-tradition / form / terminology), spend ~5-10 min reading the relevant v6.6 files.
  • When you find substrate (e.g. Switzerland Pflichtteil 2023 reform mechanics), capture the source-law citation (ZGB Art 470-480) + the structural decomposition (descendants 1/2 + spouse 1/2 + freely-disposable quota 1/2) in your scorecard / scratch notes.
  • Build v2 universal-classifier options using the captured structural decomposition. The v2 option references the SOURCE LAW citation, never the v6.6 file.
  • At step 6i RESEARCH-COMPLETENESS SELF-SCORE, RC7 scores 5/5 if all applicable v6.6 files read + distilled; 3/5 if partial; 0/5 if v6.6 not consulted on a topic where it has substrate.

Diagnostic — clean-break violations to watch for

If you find yourself writing any of the following in v2 outputs, STOP — clean-break violated:

  • ~/openinherit/code-standard/... (path reference)
  • https://openinherit.org/v3/... ($id URL reference)
  • inheritVersion: ">=3.0.0 <4.0.0" (v6.6 version-stamp)
  • v6.6 schema-stats.json says X (named-substrate-source reference)
  • per v6.6 Switzerland extension X (v6.6 module reference)
  • as in v6.6 jurisdiction-profiles.json (v6.6 file reference)

Replace with first-principles citation:

  • ZGB Art 482-484 (Swiss inheritance law)
  • Wills Act 1837 + ss.25/33
  • Succession (Scotland) Act 1964 + T&S(S)A 2024
  • Hindu Succession Act 1956 + HSA Amendment Act 2005
  • Sharia faraid jurisprudence (per academic source X)
  • etc.

Substrate inventory (read-locations only — never cite these in v2 outputs)

Top-level structure of ~/openinherit/code-standard/:

  • reference-data/jurisdiction-profiles.json (~500L) — per-jurisdiction locale + currency + professional-terminology + testamentary-freedom clause + probate-court + death-registration + typical-timeline
  • v3/extensions/ — 20 jurisdiction extensions:
    • Common-law: uk-england-wales (352L) / scotland (591L) / ireland / canada / australia-nz / hong-kong / singapore-malaysia
    • Civil-law: switzerland (179L; Pflichtteil + cantonal + matrimonial regime) / brazil / japan / prc-china / latin-america (Chilean mitad legitimaria + cuarta de mejoras)
    • Mixed: israel
    • Customary: africa-customary
    • Faith-pillars: hindu-succession (295L) / islamic-succession (421L) / jewish-succession (334L)
    • Regional: eu-succession (Brussels IV/650/2012) / uae / india
  • reference-data/event-types.json (164L) — death-related event taxonomy
  • reference-data/enum-descriptions.json (251L) — enum value descriptions
  • reference-data/form-requirements.json — form-by-jurisdiction
  • reference-data/government-submissions.json — gov submission requirements
  • reference-data/regulatory-bodies.json — regulator-by-jurisdiction
  • reference-data/gift-exemptions.json — IHT/gift-tax exemption tables
  • reference-data/funeral-providers.json — funeral provider directory
  • extensions-registry.json — master extension index

Total ~2700L of pre-authored multi-jurisdiction substrate.

Cost

~+5-10 min per substrate-rich Q. Substantially less than re-deriving the legal substrate from web research alone.

Trigger

Rich-directive 2026-05-17 BST after Q-027 retrospective identified the substrate-gap pattern: Q-026 + Q-027 first-pass E&W-creep recurrence would have been prevented if v6.6 substrate was systematically consulted. Q-027 mid-Q recovery cost ~75 min that step 6h would have eliminated.

  • [[feedback_e_w_creep_recurrence_q027_first_v3_13_production_use]] — the recurrence that fired this rule
  • [[refined_prompt_v3_14_released_wide_bundle_5_improvements_2026_05_17]] — v3.14 release (international-scope sub-audit + cascade-compliance sub-audit + closing-question boundary test + research-completeness self-score + YAML-safety)
  • brand-architecture.md v1.16+ § build-order — v6.6 deletion post-v2-release rationale
  • arch-state §A-141 standards-engagement positioning
  • Q-027 retrospective in MEMORY.md project_zeta_q27_locked_xi_eta_prime_wills_bequest_class_shape_mvp_2026_05_17

Graph View

Backlinks