Add self-scoping regression command

tests/test_cli.py

@@ -154,3 +154,100 @@ def test_compare_assessment_cli_writes_markdown_report(tmp_path):
     assert exit_code == 0
     assert "Status: `regression`" in report
     assert "Route LLM Requests Across Providers" in report
+
+
+def test_self_assess_cli_exports_challenger_and_comparison(tmp_path):
+    source = write_repo(tmp_path)
+    golden_path = tmp_path / "golden.json"
+    golden_path.write_text(
+        json.dumps(
+            {
+                "profile_id": "test-golden",
+                "ability": {
+                    "expected_capabilities": [
+                        {"name": "Expose Repository Interface"}
+                    ]
+                },
+                "forbidden_native_capabilities": [],
+            }
+        ),
+        encoding="utf-8",
+    )
+    assessment_path = tmp_path / "out" / "assessment.json"
+    comparison_path = tmp_path / "out" / "comparison.json"
+
+    exit_code = main(
+        [
+            "self-assess",
+            "--repo",
+            "Self Assess Repo",
+            "--source-path",
+            str(source),
+            "--golden",
+            str(golden_path),
+            "--assessment-output",
+            str(assessment_path),
+            "--comparison-output",
+            str(comparison_path),
+            "--format",
+            "json",
+            "--database-path",
+            str(tmp_path / "registry.sqlite3"),
+            "--checkout-root",
+            str(tmp_path / "checkouts"),
+        ]
+    )
+
+    assessment = json.loads(assessment_path.read_text(encoding="utf-8"))
+    comparison = json.loads(comparison_path.read_text(encoding="utf-8"))
+    assert exit_code == 0
+    assert assessment["target_repository"]["repo_slug"] == "self-assess-repo"
+    assert assessment["execution"]["mode"] == "deterministic-only"
+    assert comparison["status"] == "candidate_improvement"
+    assert comparison["matched_expected_capabilities"] == [
+        "Expose Repository Interface"
+    ]
+
+
+def test_self_assess_cli_can_fail_on_regression(tmp_path):
+    source = tmp_path / "provider-repo"
+    source.mkdir()
+    (source / "README.md").write_text("# Provider Repo\n", encoding="utf-8")
+    (source / "providers.py").write_text(
+        "provider_registry = {'openrouter': OpenRouterAdapter}\n",
+        encoding="utf-8",
+    )
+    golden_path = tmp_path / "golden.json"
+    golden_path.write_text(
+        json.dumps(
+            {
+                "profile_id": "test-golden",
+                "ability": {"expected_capabilities": []},
+                "forbidden_native_capabilities": [
+                    {"name": "Route LLM Requests Across Providers"}
+                ],
+            }
+        ),
+        encoding="utf-8",
+    )
+
+    exit_code = main(
+        [
+            "self-assess",
+            "--repo",
+            "Provider Repo",
+            "--source-path",
+            str(source),
+            "--golden",
+            str(golden_path),
+            "--format",
+            "json",
+            "--fail-on-regression",
+            "--database-path",
+            str(tmp_path / "registry.sqlite3"),
+            "--checkout-root",
+            str(tmp_path / "checkouts"),
+        ]
+    )
+
+    assert exit_code == 1