one time bootstrap path

tests/test_candidate_graph.py

@@ -108,6 +108,51 @@ def test_candidate_generator_enriches_descriptions_from_content_chunks():
     assert '@app.post("/classify")' in graph[0].capabilities[0].description
 
 
+def test_candidate_generator_prefers_intent_over_derived_scope_chunks():
+    repository = Repository(
+        id=1,
+        name="KeyCape",
+        url="/tmp/key-cape",
+        description=None,
+        branch="main",
+        status="analyzed",
+    )
+    facts = [
+        fact(1, "intent", "INTENT", "INTENT.md"),
+        fact(2, "scope", "SCOPE", "SCOPE.md"),
+        fact(3, "documentation", "README", "README.md"),
+    ]
+    chunks = [
+        chunk(
+            1,
+            "scope",
+            "SCOPE.md",
+            "# SCOPE\nAlready provides deployed IAM runtime behavior.",
+            end_line=2,
+        ),
+        chunk(
+            2,
+            "intent",
+            "INTENT.md",
+            "# INTENT\nDesign a lightweight IAM profile implementation.",
+            end_line=2,
+        ),
+        chunk(
+            3,
+            "documentation",
+            "README.md",
+            "# KeyCape\nREADME fallback should not beat intent.",
+            end_line=2,
+        ),
+    ]
+
+    graph = CandidateGraphGenerator().generate(repository, facts, chunks)
+
+    assert graph[0].name == "Design A Lightweight IAM Profile Implementation"
+    assert "INTENT. Design a lightweight IAM profile implementation" in graph[0].description
+    assert graph[0].source_refs[0].path == "INTENT.md"
+
+
 def test_candidate_confidence_scoring_stays_conservative_for_weak_facts():
     repository = Repository(
         id=1,