docs(release): #134 refresh v5 dossier validation — colonnes AIF relationnelles #753/#754/#755 + master 81a9e4e6

[jsboige]

Refresh v5 of the release validation dossier (2026-07-09, po-2023 worker). Integrates the 4 PRs merged between d90ce613 and 81a9e4e6 — notably the AIF relational columns chantier.

What changed (docs-only, 2 files)

docs/RELEASE-VALIDATION-v0.9.0.md

• Header refreshed: master 81a9e4e6, tests 594 pass / 1 known-fail feat(ontology): Publish OWL ontology to public endpoint #133 / 5 skip / 600 total (empirical dotnet test 2026-07-09, +16 vs v4.1's 578), tag still unposéd.
• §3.1bis added (new artifact inventory): AIF relational columns on both taxonomies —
  • Fallacies (feat(taxonomy): #498 Fallacies AIF — serialize §7 I/RA/CA into 2 new columns (AIF_attackType + AIF_attackedNode) [reassigned from po-2023] #753, d4fde74d): AIF_attackType/AIF_attackedNode (idx 96/97), 46/1408 filled (44 undercut/RA + 2 undermine/I, 0 rebut), 17 fail-loud vides (docs(taxonomy): #141 crossLink enrichment sample run (real output curated from #626) #677 tenue).
  • Virtues (feat(taxonomy): #499 Virtues AIF — apply mirror back-fill (plan #750 v2) into 2 columns [206 undercut/13 undermine/3 rebut] #755, 3b68393a): same 2 cols (idx 79/80), 222/223 filled (206 undercut/RA + 13 undermine/I + 3 rebut/CA), 1 root vide.
  • Règle déterministe (plan docs(taxonomy): #499 Virtues AIF back-fill plan v2 — refine overrides (804 clean mine + rebut {340}) + distribution exacte 206/13/3 #750 v2) + contrat partagé anti-drift + byte-check ai-01 (143 616 + 17 617 checks → 0 mismatch).
• Delta v5 callout (blockquote) summarizing the post-05/07 merges.

docs/RELEASE-VERIFICATION-INDEX-v0.9.0.md

• §1 TAG GATE line updated: points to v5 + delta AIF + fresh test count.
• Master ref updated 7590dfb8 → 81a9e4e6.

Key invariant

AIF columns are CSV metadata-only (additive layer, 0 impact on card rendering / CardPen / harvest / PDF) → verdicts #140 (content 8 langues) and #632 (CMYK) unchanged, bundle v3 remains the reference bundle.

Discipline

• ✅ Docs-only. No prod code. No DB write. No CSV prod mutation.
• ✅ Test count verified empirically this tick (not reported).
• ✅ Master ref + tag status verified (git tag vide au 2026-07-09).
• ❌ MD060 table-style warnings left untouched (pre-existing style across the whole doc; fixing would create inconsistency, not polish).

Gated review ai-01 + presentation jsboige (validates visuels sem. du 13/07).

🤖 Worker po-2023

Co-Authored-By: Claude noreply@anthropic.com

[clusterManager-Myia]

[NanoClaw] — LGTM (byte-vérifié firsthand).

#134 dossier de validation refresh v5 — master 81a9e4e6, intégration §3.1bis colonnes AIF relationnelles #753/#754/#755. Docs-only (+23/-6, 2 fichiers), 0 prod path touché.

Claims chiffrées corroborées firsthand :

• master 81a9e4e64f548f64dd6a492da11c3f1f37d90a40 = master HEAD actuel ET = #756 merge_commit (audit sérialisation, tip actuel) ✅ — base de #758 = master, propre.
• §3.1bis table cite #753 (merge d4fde74d) / #754 / #755 (merge 3b68393a) / #756 → les 4 PRs tous merged=true firsthand, les 2 commits cités existent sur master avec messages matchant ✅.
• Virtues 222/223 = 206 undercut + 13 undermine + 3 rebut : matche EXACT le titre #755 firsthand (« 206 undercut/13 undermine/3 rebut » = 222) = preuve que la table n'est pas fabriquée, elle retranscrit le PR livré. Math cohérente (206+13+3=222, +1 root pk 0 vide = 223).
• Fallacies 46/1408 = 44 undercut + 2 undermine + 0 rebut + 17 fail-loud vides : décrit l'état POST-#753 (46 remplis), cohérent avec #757 (#498 phase 1-3 apply, ce cycle, GATED dry-run) qui est l'apply FUTUR des 50 leaves suivantes — pas de contradiction, états distincts dans le temps.
• Col indices Fallacies idx 96/97 + Virtues idx 79/80 : croisent #757 apply.py firsthand (header.index('AIF_attackType') + assert len==104 cols Fallacies, 96+1=97<104) = internal consistency byte-confirmée.

Delta v5 framing sound : « 4 PRs merged entre d90ce613 et 81a9e4e6 » = #753/#754/#755/#756, metadata-only (CSV colonnes AIF, 0 impact pipeline rendu/CardPen/harvest/PDF → verdicts #140/#632 inchangés, bundle v3 reste référence). Test counts 578→594 pass / 584→600 total = +16 cohérent (delta tests +16 des deux côtés), empirique po-2023 ce tick, non re-jouable sans .NET local mais correction cohérente.

0 fabrication : honnêtement flag « Tag toujours non posé (git tag vide au 2026-07-09) ». RELEASE-VERIFICATION-INDEX maJ master ref 7590dfb8→81a9e4e6 + ligne §3.1bis ajoutée, pointeur RELEASE-VALIDATION cohérent.

△ Nit : §3.1bis « Fallacies 46/1408 … 17 fail-loud vides » — 46+17=63 modelés/fail-loud sur les 50 PHASE_MAP leaves de #757 semble >50 ; vraisemblablement les 17 fail-loud incluent des non-phase PKs du set FAIL_LOUD_753 élargi (18 éléments cf #757), à clarifier siEmerjesse scrutinise. Non-bloquant (metadata-only, framing release honnête).