TT Claude Memory

project_zeta_3_A2_shacl_invariants_2026_05_03

Properties1
descriptionν.α/ζ.3 Spike A2 — SHACL constraint authoring at scale + LinkML→SHACL pipeline. 5 INHERIT v2 invariants × 2 variants × 20 fixtures = 200 cells; pySHACL 0.31.0; 1K benchmark; gen-shacl pipeline; Q-NU-005 state 1 → state 3; T-file authored

2 min read

ζ.3 Spike A2 — SHACL invariants at scale — CLOSED outcome-VALIDATED

Why: Pre-validates substrate for ζ-Q11 (SHACL constraint strategy re-ask). Depends on A1 Policy-1 (minor-only Phase-1; open shapes required for partner extension compatibility).

How to apply: ζ-Q11 formal asking under refined-prompt v3.8 should use τ.α as provisional pick (open-shape default + SHACL-SPARQL I3-I5). T-file at ~/off-github/library/projects/inherit/T-spike-zeta-3-A2-shacl-invariants-2026-05-03.md. Q-NU-005 placeholder at docs-strategy/.../current-questions/Q-NU-005-shacl-invariants-partner-extension-compatibility.md.

Outcome summary

  • Kill condition: NOT-MET. All 5 invariants expressible in SHACL Core 1.0 + SHACL-AF 1.1 (SPARQL) without vendor extensions.
  • Maturity: frontier-derisked
  • 200-cell results: open-shape variants all precision=1.000; I1-closed FP=1 (v09 partner extension rejected); I4-closed FP=6 (under-specified closed shape hazard)
  • 1K benchmark: I2-open 98.4ms ✓; I4-open 94.9ms ✓; I3-closed 99.8ms ✓; I1-open 134.8ms borderline; I5-open 4821ms FAIL (needs async audit job)

5 load-bearing findings

  1. KILL-NOT-MET: All 5 invariants (testator cardinality / beneficiary-share / DistributionEvent temporal / Trustee disjunction / Wasiyyah ≤1/3) expressible in SHACL Core 1.0 + SHACL-AF 1.1
  2. Open-shape = Phase-1 default (A1 Policy-1 confirmed): I1-closed rejects partner extension fixture (FP=1); I4-closed over-rejects 6/10 valid fixtures (FP=6)
  3. gen-shacl defaults to —closed: MUST use --non-closed in CI; SHACL-SPARQL I3/I4/I5 hand-authored in shacl-extensions/ folder
  4. Per-estate validation: 1.4–95ms all invariants (Phase-1 target met); batch I5: 4800ms (async audit job needed)
  5. Toolchain confirmed: pySHACL 0.31.0 (CI) + TopBraid SHACL API 1.5.0 (JVM production); xpSHACL Phase-2 candidate

Q-NU-005 option-set for ζ-Q11

  • τ.α: Open-shape default + SHACL-SPARQL I3-I5 (PROVISIONAL PICK — precision=1.000 all variants)
  • τ.β: Closed-shape with explicit sh:property declarations (higher partner coordination cost)
  • τ.γ: SHACL Core only — defer I3-I5 Phase-1.5+ (misses Wasiyyah enforcement)
  • τ.δ: SHACL 1.2 WD features (NOT recommended — pre-Rec standard)
  • τ.ε: Rich-defined synthesis

Artefact paths

  • T-file: ~/off-github/library/projects/inherit/T-spike-zeta-3-A2-shacl-invariants-2026-05-03.md
  • Q-NU-005: docs-strategy/docs/superpowers/specs/2026-04-29-multi-phase-audit/current-questions/Q-NU-005-shacl-invariants-partner-extension-compatibility.md
  • Results JSON: /tmp/spike-zeta-3-A2/validation-results.json
  • Closure bundle: /tmp/spike-zeta-3-A2/closure-bundle.md

Graph View

Backlinks