refactor(llm): elevate prompt engineering with context grounding and contrastive taxonomy

graphify-out/2026-08-21/.graphify_labels.json

@@ -45,7 +45,7 @@
   "43": "2. Basic CLI Usage Examples",
   "44": "2. Standard Streams & Exit Codes",
   "45": "ClassificationResult",
-  "46": "InherenceClassifier",
+  "46": "test_adversarial.py",
   "47": "classifier.py",
   "48": "test_convert_article_to_markdown.py",
   "49": "content_northvolt_de.md",
@@ -151,7 +151,7 @@
   "149": "get_hl_gl_ceid",
   "150": "13. Estratégia de testes",
   "151": "6. Contrato de entrada",
-  "152": "assemble_markdown_document",
+  "152": "InherenceClassifier",
   "153": "convert_html_to_markdown",
   "154": "JSON Schema Contract: Deterministic Article Content Selection",
   "155": "5. Escopo",
@@ -163,5 +163,8 @@
   "161": "test_normalize_list_deduplication_preserves_case_and_order",
   "162": "test_normalize_date_iso_8601_variants",
   "163": "test_metadata_priority_original_url_all_fallbacks",
-  "164": "test_normalize_scalar_non_string_types"
+  "164": "test_normalize_scalar_non_string_types",
+  "165": "LLMFallbackAdapter",
+  "166": "remove_duplicate_initial_h1",
+  "167": "test_normalize_scalar_whitespace_collapsing"
 }