baseline repo characteristics no longer crowd the candidate graph

src/repo_registry/candidate_graph/generator.py

@@ -100,10 +100,6 @@ class CandidateGraphGenerator:
         )
 
         capabilities: list[CandidateCapabilityDraft] = []
-        if interfaces:
-            capabilities.append(
-                self._interface_capability(interfaces, tests, examples, docs, chunks)
-            )
         capabilities.extend(
             self._intent_capabilities(intent_facts, chunks, tests, examples, docs)
         )
@@ -127,31 +123,9 @@ class CandidateGraphGenerator:
                     docs,
                 )
             )
-        if manifests or frameworks or languages:
+        if interfaces and not capabilities:
             capabilities.append(
-                CandidateCapabilityDraft(
-                    name="Describe Repository Structure",
-                    description=(
-                        "Summarize detected languages, package manifests, and framework "
-                        "hints as structural context for review."
-                    ),
-                    inputs=[],
-                    outputs=["repository structure summary"],
-                    confidence=self._structure_confidence(
-                        manifests=manifests,
-                        frameworks=frameworks,
-                        languages=languages,
-                        docs=docs,
-                    ),
-            source_refs=self._source_refs(manifests + frameworks + languages),
-            primary_class="repository-structure",
-            attributes=self._structure_attributes(
-                manifests,
-                frameworks,
-                        languages,
-                    ),
-                    evidence=self._evidence(tests, examples, docs),
-                )
+                self._interface_capability(interfaces, tests, examples, docs, chunks)
             )
 
         return [
@@ -356,7 +330,10 @@ class CandidateGraphGenerator:
                     continue
                 if line.startswith("#"):
                     heading = line.lstrip("#").strip().lower()
-                    in_capability_section = "capabilit" in heading
+                    in_capability_section = (
+                        "capabilit" in heading
+                        or heading in {"primary utility", "core utility"}
+                    )
                     continue
                 if not in_capability_section:
                     continue
@@ -367,11 +344,23 @@ class CandidateGraphGenerator:
         return items
 
     def _intent_capability_name(self, text: str) -> str:
+        lowered = re.sub(r"[*_`]", "", text.lower())
+        if "continuous connectivity" in lowered and "remote systems" in lowered:
+            return "Maintain Continuous Connectivity Between Remote Systems And Central Hub"
+        if "observable" in lowered and "auditable" in lowered and "controllable" in lowered:
+            return "Make Connectivity Observable Auditable And Controllable"
+        if "cli tool" in lowered and "mcp" in lowered:
+            return "Expose CLI And MCP Accessible Service"
         candidate = re.split(r"\s+-\s+|\s*:\s*|[.!?]\s+", text.strip(), maxsplit=1)[0]
         candidate = candidate.strip(" .:-")
         if not candidate:
             return ""
-        return self._title_from_words(candidate.split()[:8])
+        words = candidate.split()
+        if words:
+            words[0] = self._imperative_verb(words[0])
+        while words and words[-1].lower().strip(",;:") in {"a", "an", "the", "and", "or", "as", "both"}:
+            words.pop()
+        return self._title_from_words(words[:10])
 
     def _interface_features(
         self,
@@ -508,16 +497,36 @@ class CandidateGraphGenerator:
             [
                 repository.name,
                 repository.description or "",
-                " ".join(chunk.text[:600] for chunk in chunks if chunk.kind == "documentation"),
-                " ".join(f"{fact.kind} {fact.name} {fact.value}" for fact in facts),
+                " ".join(
+                    chunk.text[:600]
+                    for chunk in chunks
+                    if chunk.kind in {"intent", "documentation"}
+                    and chunk.metadata.get("source_role") != "agent_guidance"
+                ),
+                " ".join(
+                    f"{fact.kind} {fact.name} {fact.value}"
+                    for fact in facts
+                    if not (
+                        fact.kind == "llm_provider"
+                        and self._utility_relationship(fact) not in {"owned", "facade", "adapter"}
+                    )
+                ),
             ]
         ).lower()
         attributes: list[str] = []
-        if any(token in text for token in ("repository", "repo", "registry")):
-            attributes.append("repository")
+        if any(token in text for token in ("ssh", "tunnel", "reverse tunnel", "remote access", "connectivity")):
+            attributes.extend(["remote-access", "connectivity"])
+            if any(token in text for token in ("audit", "health check", "lifecycle", "ops", "operator")):
+                attributes.append("operations")
+            return "it-operations", self._unique(attributes)
         if any(token in text for token in ("ability", "capability", "feature")):
             return "repository-intelligence", self._unique(attributes + ["capability-mapping"])
-        if any(token in text for token in ("llm", "openrouter", "claude", "model provider")):
+        promotable_llm = any(
+            fact.kind == "llm_provider"
+            and self._utility_relationship(fact) in {"owned", "facade", "adapter"}
+            for fact in facts
+        )
+        if promotable_llm:
             return "ai-integration", self._unique(attributes + ["llm-provider"])
         if any(fact.kind == "interface" for fact in facts):
             attributes.append("interface")
@@ -777,6 +786,9 @@ class CandidateGraphGenerator:
         repository: Repository,
         chunks: list[ContentChunk],
     ) -> str:
+        ops_name = self._operations_ability_name(chunks)
+        if ops_name:
+            return ops_name
         purpose_text = self._document_purpose_sentence(chunks) or repository.description
         if purpose_text:
             normalized = self._imperative_purpose(purpose_text)
@@ -794,9 +806,24 @@ class CandidateGraphGenerator:
                 return paragraph
         return ""
 
+    def _operations_ability_name(self, chunks: list[ContentChunk]) -> str:
+        text = " ".join(
+            chunk.text
+            for chunk in self._documentation_chunks(chunks)
+            if chunk.kind == "intent"
+        ).lower()
+        if "ssh reverse tunnel" in text or "ssh reverse tunneling" in text:
+            return "Manage SSH Reverse Tunnel Connectivity"
+        return ""
+
     def _imperative_purpose(self, text: str) -> str:
         cleaned = re.sub(r"\s+", " ", text.strip())
         cleaned = re.split(r"[.!?]\s+", cleaned, maxsplit=1)[0]
+        cleaned = re.sub(
+            r"(?i)^this\s+repository\s+exists\s+to\s+provide\s+(?:an?\s+)?",
+            "Provide ",
+            cleaned,
+        )
         cleaned = re.sub(r"^[A-Z][A-Za-z0-9_-]*\s+(?:is|provides|offers)\s+", "", cleaned)
         cleaned = cleaned.strip(" .:-")
         if not cleaned:
@@ -816,6 +843,8 @@ class CandidateGraphGenerator:
         }
         if lower in irregular:
             return irregular[lower]
+        if lower in {"this"}:
+            return lower
         if lower.endswith("ies") and len(lower) > 4:
             return f"{lower[:-3]}y"
         if lower.endswith(("des", "ses", "tes", "ves", "zes")) and len(lower) > 4:

src/repo_registry/repo_scanning/scanner.py

@@ -474,7 +474,11 @@ class DeterministicScanner:
             return "ci_tooling"
         if lower.startswith(("tests/", "test/")) or name.startswith("test_"):
             return "test_evidence"
-        if name.startswith("readme") or lower.startswith(("docs/", "doc/", "wiki/")):
+        if (
+            name.startswith("readme")
+            or name.endswith(".md")
+            or lower.startswith(("docs/", "doc/", "wiki/", "workplans/", "architecture/"))
+        ):
             return "product_documentation"
         if name in MANIFEST_FRAMEWORK_HINTS or name.endswith((".lock", ".mod")):
             return "dependency_declaration"
@@ -483,13 +487,21 @@ class DeterministicScanner:
         return "implementation_source"
 
     def _has_provider_signal(self, lower_text: str, needle: str) -> bool:
-        pattern = re.compile(rf"(?<![a-z0-9-]){re.escape(needle.lower())}(?![a-z0-9-])")
+        if f"{needle.lower()}_api_key" in lower_text:
+            return True
+        pattern = re.compile(rf"(?<![a-z0-9_-]){re.escape(needle.lower())}(?![a-z0-9_-])")
         for match in pattern.finditer(lower_text):
             context = lower_text[max(0, match.start() - 20) : match.end() + 20]
             if needle == "claude" and (
                 "claude.md" in context
                 or "claude code" in context
                 or "claude.ai/code" in context
+                or "claude mcp" in context
+                or "mcp" in context
+                or ".claude" in context
+                or "claude.json" in context
+                or "claude plugin" in context
+                or "claude prompt" in context
             ):
                 continue
             return True

src/repo_registry/web_ui/views.py

@@ -541,6 +541,19 @@ def render_analysis_diagnostics(
                 ),
             )
         )
+    elif capability_count == 0:
+        notices.append(
+            (
+                "warn",
+                "No domain capabilities were produced.",
+                (
+                    "The scanner found repository evidence, but only baseline "
+                    "context or weak documentation was available. If this "
+                    "repository should provide concrete capabilities, record an "
+                    "expectation gap for the missing behavior."
+                ),
+            )
+        )
     elif only_weak_candidates:
         notices.append(
             (

tests/fixtures.py

@@ -163,3 +163,62 @@ def write_dependency_only_repo(root: Path) -> Path:
         encoding="utf-8",
     )
     return repo
+
+
+def write_ops_bridge_like_repo(root: Path) -> Path:
+    repo = root / "ops-bridge-like"
+    repo.mkdir()
+    (repo / "INTENT.md").write_text(
+        "# INTENT\n\n"
+        "## Purpose\n\n"
+        "This repository exists to provide a **reliable, inspectable, and controllable "
+        "connectivity layer** between distributed dev, build, test and execution "
+        "environments for dev and ops personal human and agentic.\n\n"
+        "## Primary Utility\n\n"
+        "The repository provides a **managed SSH reverse tunneling system** that:\n\n"
+        "* Maintains continuous connectivity between remote systems and a central hub\n"
+        "* Makes connectivity **observable, auditable, and controllable**\n"
+        "* Exposes this capability as both a **CLI tool and an MCP-accessible service**\n\n"
+        "## Intended Users\n\n"
+        "* Human operators managing infrastructure and connectivity\n"
+        "* LLM-based agents requiring stable access to local services\n",
+        encoding="utf-8",
+    )
+    (repo / "SCOPE.md").write_text(
+        "# SCOPE\n\nSSH reverse tunnel lifecycle manager for remote execution environments.\n",
+        encoding="utf-8",
+    )
+    (repo / "README.txt").write_text(
+        "# Ops Bridge\n\n"
+        "Manages named SSH reverse tunnels with auto-reconnect, health checks, "
+        "audit logging, and an MCP server so Claude Code can start and inspect tunnels.\n",
+        encoding="utf-8",
+    )
+    (repo / "pyproject.toml").write_text(
+        "[project]\ndependencies = ['typer', 'pytest']\n",
+        encoding="utf-8",
+    )
+    (repo / "scripts").mkdir()
+    (repo / "scripts" / "register_mcp.py").write_text(
+        '"""Register the ops-bridge MCP server with Claude MCP."""\n'
+        "from pathlib import Path\n"
+        "CLAUDE_JSON = Path.home() / '.claude.json'\n"
+        "def main():\n"
+        "    return CLAUDE_JSON.exists()\n",
+        encoding="utf-8",
+    )
+    (repo / "bridge").mkdir()
+    (repo / "bridge" / "cli.py").write_text(
+        "import typer\n"
+        "app = typer.Typer()\n"
+        "@app.command()\n"
+        "def up(name: str):\n"
+        "    return name\n",
+        encoding="utf-8",
+    )
+    (repo / "workplans").mkdir()
+    (repo / "workplans" / "BRIDGE-WP-0003.md").write_text(
+        "# MCP skill work\n\nClaude Code sessions can call bridge_status().\n",
+        encoding="utf-8",
+    )
+    return repo

tests/test_candidate_graph.py

@@ -68,12 +68,7 @@ def test_candidate_generator_builds_purpose_seed_from_observed_facts():
     assert interface_capability.features[0].name == "POST /classify"
     assert interface_capability.features[0].location == "app.py"
     assert interface_capability.evidence[0].strength == "strong"
-    structure_capability = ability.capabilities[1]
-    assert structure_capability.name == "Describe Repository Structure"
-    assert {
-        "utility-dependency",
-        "review-required-structural-context",
-    } <= set(structure_capability.attributes)
+    assert len(ability.capabilities) == 1
 
 
 def test_candidate_generator_extracts_intended_capability_blocks_from_intent_chunks():
@@ -265,7 +260,7 @@ def test_candidate_confidence_scoring_increases_with_supporting_facts():
 
     assert graph[0].confidence == 1.0
     assert graph[0].capabilities[0].confidence == 0.85
-    assert graph[0].capabilities[1].confidence == 0.75
+    assert len(graph[0].capabilities) == 1
 
 
 def test_candidate_generator_names_cli_features_from_nearby_function():

tests/test_registry_service.py

@@ -15,6 +15,7 @@ from tests.fixtures import (
     write_javascript_typescript_package_repo,
     write_key_cape_like_repo,
     write_llm_connect_like_repo,
+    write_ops_bridge_like_repo,
     write_misleading_docs_repo,
     write_python_cli_repo,
     write_readme_only_repo,
@@ -27,6 +28,32 @@ def make_service(tmp_path):
     return RegistryService(store, ingestion=GitIngestionService(tmp_path / "checkouts"))
 
 
+def add_candidate_capability(service, repository_id, analysis_run_id, ability_id, name):
+    with service.store.connect() as connection:
+        cursor = connection.execute(
+            """
+            INSERT INTO candidate_capabilities
+              (repository_id, analysis_run_id, ability_id, name, description,
+               inputs, outputs, primary_class, attributes, confidence, source_refs)
+            VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """,
+            (
+                repository_id,
+                analysis_run_id,
+                ability_id,
+                name,
+                "Review target capability inserted for review workflow tests.",
+                "[]",
+                "[]",
+                "test-capability",
+                json.dumps(["test-review-target"]),
+                0.5,
+                "[]",
+            ),
+        )
+        return int(cursor.lastrowid)
+
+
 class FakeLLMExtractor:
     def __init__(self, abilities):
         self.abilities = abilities
@@ -298,12 +325,12 @@ def test_search_filters_by_status_language_and_framework(tmp_path):
     service.approve_candidate_graph(repository.id, summary.analysis_run.id)
 
     results = service.search(
-        "repository",
+        "health",
         status="indexed",
         language="Python",
         framework="FastAPI",
         ability="Support Filterable",
-        capability="Repository Structure",
+        capability="Repository Interface",
     )
     wrong_language_results = service.search(
         "repository",
@@ -380,7 +407,6 @@ def test_fixture_breadth_javascript_typescript_package_extracts_structure_and_ap
     assert ("framework", "React", "package.json") in fact_names
     assert ("framework", "Vitest", "package.json") in fact_names
     assert "Expose Repository Interface" in capability_names
-    assert "Describe Repository Structure" in capability_names
     assert "API" in feature_types
     assert service.ability_map(repository.id).abilities == []
 
@@ -468,10 +494,36 @@ def test_regression_dependency_only_repo_keeps_libraries_as_context(tmp_path):
         for capability in ability.capabilities
     }
     assert "Route LLM Requests Across Providers" not in capability_names
-    assert capability_names == {"Describe Repository Structure"}
-    structure = graph.abilities[0].capabilities[0]
-    assert "utility-dependency" in structure.attributes
-    assert "review-required-structural-context" in structure.attributes
+    assert capability_names == set()
+    assert any(fact.kind == "manifest" for fact in summary.facts)
+
+
+def test_regression_ops_bridge_like_repo_is_it_operations_not_llm_provider(tmp_path):
+    source = write_ops_bridge_like_repo(tmp_path)
+    service = make_service(tmp_path)
+    repository = service.register_repository(name="Ops Bridge Marketing Name", url=str(source))
+
+    summary = service.analyze_repository(repository.id, use_llm_assistance=False)
+    graph = service.candidate_graph(repository.id, summary.analysis_run.id)
+
+    ability = graph.abilities[0]
+    capability_names = {
+        capability.name
+        for candidate_ability in graph.abilities
+        for capability in candidate_ability.capabilities
+    }
+    facts = {(fact.kind, fact.name, fact.path) for fact in summary.facts}
+    assert ability.name == "Manage SSH Reverse Tunnel Connectivity"
+    assert ability.primary_class == "it-operations"
+    assert {"remote-access", "connectivity", "operations"} <= set(ability.attributes)
+    assert "repository" not in ability.attributes
+    assert "llm-provider" not in ability.attributes
+    assert "Route LLM Requests Across Providers" not in capability_names
+    assert "Maintain Continuous Connectivity Between Remote Systems And Central Hub" in capability_names
+    assert "Make Connectivity Observable Auditable And Controllable" in capability_names
+    assert "Expose CLI And MCP Accessible Service" in capability_names
+    assert ("llm_provider", "Claude", "scripts/register_mcp.py") not in facts
+    assert ("llm_provider", "Claude", "workplans/BRIDGE-WP-0003.md") not in facts
 
 
 def test_fixture_breadth_empty_repo_produces_no_candidate_claims(tmp_path):
@@ -933,20 +985,14 @@ def test_analyze_repository_can_trusted_auto_approve_candidates(tmp_path):
         for capability in graph.abilities[0].capabilities
     }
     assert statuses_by_capability["Expose Repository Interface"] == "approved"
-    assert statuses_by_capability["Describe Repository Structure"] == "candidate"
     assert ability_map.abilities[0].name == "Report Health Over HTTP"
     assert decisions[0].action == "trusted_auto_approve_candidate_graph"
     assert "deterministic candidate generation" in decisions[0].notes
-    assert "Auto-approved 1 safe candidate capability(s); left 1 for review." in decisions[0].notes
+    assert "Auto-approved 1 safe candidate capability(s); left 0 for review." in decisions[0].notes
     assert (
         "Approved: Expose Repository Interface: owned interface with sufficient confidence."
         in decisions[0].notes
     )
-    assert (
-        "Skipped: Describe Repository Structure: structural/dependency context "
-        "requires curator review."
-        in decisions[0].notes
-    )
 
 
 def test_rebuild_characteristics_dry_run_preserves_approved_map(tmp_path):
@@ -1523,6 +1569,14 @@ def test_relink_candidate_feature_and_evidence_to_another_capability(tmp_path):
     repository = service.register_repository(name="Relink Leaves", url=str(source))
     summary = service.analyze_repository(repository.id)
     graph = service.candidate_graph(repository.id, summary.analysis_run.id)
+    add_candidate_capability(
+        service,
+        repository.id,
+        summary.analysis_run.id,
+        graph.abilities[0].id,
+        "Review Target Capability",
+    )
+    graph = service.candidate_graph(repository.id, summary.analysis_run.id)
     source_capability = graph.abilities[0].capabilities[0]
     target_capability = graph.abilities[0].capabilities[1]
     feature = source_capability.features[0]
@@ -1625,6 +1679,14 @@ def test_merge_candidate_capability_moves_children_to_target(tmp_path):
     repository = service.register_repository(name="Merge Capability", url=str(source))
     summary = service.analyze_repository(repository.id)
     graph = service.candidate_graph(repository.id, summary.analysis_run.id)
+    add_candidate_capability(
+        service,
+        repository.id,
+        summary.analysis_run.id,
+        graph.abilities[0].id,
+        "Review Target Capability",
+    )
+    graph = service.candidate_graph(repository.id, summary.analysis_run.id)
     source_capability = graph.abilities[0].capabilities[0]
     target_capability = graph.abilities[0].capabilities[1]
 

tests/test_web_api.py

@@ -1,9 +1,47 @@
+import json
+import sqlite3
+
 from fastapi.testclient import TestClient
 
 from repo_registry.web_api import app as app_module
 from repo_registry.web_api.app import Settings, app, get_service, get_settings
 
 
+def add_candidate_capability(database_path, repository_id, analysis_run_id, name):
+    with sqlite3.connect(database_path) as connection:
+        ability_id = connection.execute(
+            """
+            SELECT id FROM candidate_abilities
+            WHERE repository_id = ? AND analysis_run_id = ?
+            ORDER BY id
+            LIMIT 1
+            """,
+            (repository_id, analysis_run_id),
+        ).fetchone()[0]
+        cursor = connection.execute(
+            """
+            INSERT INTO candidate_capabilities
+              (repository_id, analysis_run_id, ability_id, name, description,
+               inputs, outputs, primary_class, attributes, confidence, source_refs)
+            VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """,
+            (
+                repository_id,
+                analysis_run_id,
+                ability_id,
+                name,
+                "Review target capability inserted for API review workflow tests.",
+                "[]",
+                "[]",
+                "test-capability",
+                json.dumps(["test-review-target"]),
+                0.5,
+                "[]",
+            ),
+        )
+        return int(cursor.lastrowid)
+
+
 def test_openapi_groups_agent_facing_endpoints():
     client = TestClient(app)
 
@@ -875,10 +913,11 @@ def test_api_analysis_run_loop(tmp_path):
         '{"dependencies":{"react":"latest","vite":"latest"}}',
         encoding="utf-8",
     )
+    database_path = str(tmp_path / "api-analysis.sqlite3")
 
     def override_settings():
         return Settings(
-            database_path=str(tmp_path / "api-analysis.sqlite3"),
+            database_path=database_path,
             checkout_root=str(tmp_path / "api-checkouts"),
         )
 
@@ -903,6 +942,12 @@ def test_api_analysis_run_loop(tmp_path):
         assert get_run_response.status_code == 200
         assert get_run_response.json()["id"] == run["analysis_run"]["id"]
 
+        add_candidate_capability(
+            database_path,
+            repository_id,
+            run["analysis_run"]["id"],
+            "Describe Frontend Stack",
+        )
         candidate_response = client.get(
             f"/repos/{repository_id}/analysis-runs/"
             f"{run['analysis_run']['id']}/candidate-graph"
@@ -954,6 +999,12 @@ def test_api_analysis_run_loop(tmp_path):
         run_response = client.post(f"/repos/{repository_id}/analysis-runs", json={})
         assert run_response.status_code == 201
         run = run_response.json()
+        add_candidate_capability(
+            database_path,
+            repository_id,
+            run["analysis_run"]["id"],
+            "Describe Frontend Stack",
+        )
         candidate_response = client.get(
             f"/repos/{repository_id}/analysis-runs/"
             f"{run['analysis_run']['id']}/candidate-graph"
@@ -1358,7 +1409,7 @@ def test_ui_register_analyze_and_approve_loop(tmp_path):
         assert "Analysis completed with reviewable results." in run_detail.text
         assert "Candidate Graph" in run_detail.text
         assert "1 abilities" in run_detail.text
-        assert "2 capabilities" in run_detail.text
+        assert "1 capabilities" in run_detail.text
         assert "2 features" in run_detail.text
         assert "8 facts" in run_detail.text
         assert "Content Chunks" in run_detail.text
@@ -1426,11 +1477,11 @@ def test_ui_register_analyze_and_approve_loop(tmp_path):
         assert "1 scope" in approved_detail.text
         assert "supports" in approved_detail.text
         assert "1 abilities" in approved_detail.text
-        assert "2 capabilities" in approved_detail.text
+        assert "1 capabilities" in approved_detail.text
         assert "2 features" in approved_detail.text
         assert "Latest Candidate Graph" in approved_detail.text
         assert "1 candidate abilities" in approved_detail.text
-        assert "2 candidate capabilities" in approved_detail.text
+        assert "1 candidate capabilities" in approved_detail.text
         assert "2 candidate features" in approved_detail.text
         assert "8 candidate facts" in approved_detail.text
         assert "Use Approved Registry" in approved_detail.text
@@ -1801,6 +1852,44 @@ def test_ui_analysis_run_diagnostics_explain_failures_and_empty_results(tmp_path
         app.dependency_overrides.clear()
 
 
+def test_ui_analysis_run_diagnostics_warn_when_only_baseline_context_exists(tmp_path):
+    source = tmp_path / "dependency-only-ui"
+    source.mkdir()
+    (source / "README.md").write_text("# Dependency Only\nUses libraries.\n", encoding="utf-8")
+    (source / "requirements.txt").write_text("fastapi\npytest\n", encoding="utf-8")
+
+    def override_settings():
+        return Settings(
+            database_path=str(tmp_path / "ui-baseline-diagnostics.sqlite3"),
+            checkout_root=str(tmp_path / "ui-baseline-diagnostics-checkouts"),
+        )
+
+    app.dependency_overrides[get_settings] = override_settings
+    client = TestClient(app)
+    try:
+        repository = client.post(
+            "/repos",
+            json={
+                "url": str(source),
+                "name": "Dependency Only UI",
+                "description": "Used for baseline diagnostics.",
+            },
+        ).json()
+        run = client.post(
+            f"/ui/repos/{repository['id']}/analysis-runs",
+            data={"source_path": "", "use_llm_assistance": ""},
+            follow_redirects=False,
+        )
+
+        detail = client.get(run.headers["location"])
+
+        assert detail.status_code == 200
+        assert "No domain capabilities were produced." in detail.text
+        assert "only baseline context or weak documentation was available" in detail.text
+    finally:
+        app.dependency_overrides.clear()
+
+
 def test_ui_register_and_explore_lands_on_analysis_result(tmp_path):
     source = tmp_path / "explore-repo"
     source.mkdir()
@@ -2550,9 +2639,11 @@ def test_api_relinks_candidate_feature_and_evidence(tmp_path):
         "    return {}\n",
         encoding="utf-8",
     )
+    database_path = str(tmp_path / "api-relink.sqlite3")
+
     def override_settings():
         return Settings(
-            database_path=str(tmp_path / "api-relink.sqlite3"),
+            database_path=database_path,
             checkout_root=str(tmp_path / "api-relink-checkouts"),
         )
 
@@ -2566,6 +2657,12 @@ def test_api_relinks_candidate_feature_and_evidence(tmp_path):
         repository_id = repository_response.json()["id"]
         run_response = client.post(f"/repos/{repository_id}/analysis-runs", json={})
         run_id = run_response.json()["analysis_run"]["id"]
+        add_candidate_capability(
+            database_path,
+            repository_id,
+            run_id,
+            "Review Target Capability",
+        )
         graph_response = client.get(
             f"/repos/{repository_id}/analysis-runs/{run_id}/candidate-graph"
         )
@@ -2631,10 +2728,11 @@ def test_api_merges_candidate_capability_feature_and_evidence(tmp_path):
         "    click.echo('ok')\n",
         encoding="utf-8",
     )
+    database_path = str(tmp_path / "api-merge.sqlite3")
 
     def override_settings():
         return Settings(
-            database_path=str(tmp_path / "api-merge.sqlite3"),
+            database_path=database_path,
             checkout_root=str(tmp_path / "api-merge-checkouts"),
         )
 
@@ -2648,6 +2746,12 @@ def test_api_merges_candidate_capability_feature_and_evidence(tmp_path):
         repository_id = repository_response.json()["id"]
         run_response = client.post(f"/repos/{repository_id}/analysis-runs", json={})
         run_id = run_response.json()["analysis_run"]["id"]
+        add_candidate_capability(
+            database_path,
+            repository_id,
+            run_id,
+            "Review Target Capability",
+        )
         graph_response = client.get(
             f"/repos/{repository_id}/analysis-runs/{run_id}/candidate-graph"
         )