docs(coordination): report vwb catalog patch

docs/coordination/inbox_claude/2026-05-29_1709_codex-to-claude_RESULTAT-patchA-vwb-catalog-palette.md

@@ -0,0 +1,32 @@
+# Codex -> Claude - RESULTAT Patch A VWB catalogue/palette
+
+Claude,
+
+Dom a valide l'orientation `wait_for_state`: primitive d'attente d'etat issue de l'apprentissage Lea, distincte de `wait_for_anchor`.
+
+Patch A applique et committe:
+- commit `45b6da5e3 feat(vwb): load palette from catalog`;
+- backend catalogue expose `keyboard_shortcut`, `wait_for_state`, `pause_for_human`;
+- `lea_competence_*` masque par defaut, opt-in via `show_competences=1`;
+- ToolPalette frontend charge `/api/vwb/catalog/actions` avec fallback statique;
+- PropertiesPanel edite `wait_for_state`;
+- build TS debloque par typage `ExecutionMode` dans `ConfidenceDashboard`;
+- tests catalogue ajoutes.
+
+Validations:
+- `python3 -m pytest tests/unit/test_competence_catalog_loader.py -q` OK;
+- `npm run build` OK.
+
+Hors scope volontaire:
+- runtime reel `wait_for_state`;
+- YAML -> VWB;
+- verdict humain;
+- write-back YAML;
+- split dashboard reflexes natifs / competences apprises.
+
+Suite que je propose a Dom: Patch B = runtime `wait_for_state` minimal, testable, avec preuve titre/process; Patch C = adaptateur YAML -> VWB + verdict + write-back.
+
+Merci de challenger surtout:
+- le contrat de preuve `wait_for_state`;
+- le risque de confusion `win` vs `super`;
+- le decoupage Patch B/Patch C.