diff --git a/tests/test_kin_109_regression.py b/tests/test_kin_109_regression.py
index 519ab4b..0049649 100644
--- a/tests/test_kin_109_regression.py
+++ b/tests/test_kin_109_regression.py
@@ -5,14 +5,27 @@ Root cause: Makefile test target used bare `pytest tests/` which resolves to the
 system pytest (Python 3.14). Tests written for Python 3.11 could fail under 3.14.
 
 Fix: Makefile test target changed to `python3.11 -m pytest tests/`.
+PR fix: _detect_test_command() now returns sys.executable -m pytest for
+        pyproject.toml/setup.py projects (was bare 'pytest').
 
 Coverage:
 (1) Makefile test target uses python3.11, not bare pytest
 (2) _detect_test_command still returns 'make test' for projects with Makefile
+(3) _detect_test_command returns sys.executable -m pytest for pyproject.toml/setup.py
+(4) Pipeline auto-test uses project.test_command when explicitly set
+(5) Pipeline auto-test uses _detect_test_command when project.test_command is NULL
+(6) PATCH /api/projects: test_command absent → DB field unchanged (decision #580)
+(7) Regression: pipeline auto-test early return (no test framework) does not crash
 """
 
+import sys
 import re
 from pathlib import Path
+from unittest.mock import patch, MagicMock
+
+import pytest
+from core.db import init_db
+from core import models
 
 
 # ---------------------------------------------------------------------------
@@ -73,3 +86,177 @@ class TestDetectTestCommandUnchanged:
         assert result == "make test", (
             f"kin project should detect 'make test', got {result!r}"
         )
+
+
+# ---------------------------------------------------------------------------
+# (3) _detect_test_command: pyproject.toml/setup.py → sys.executable -m pytest
+# ---------------------------------------------------------------------------
+
+class TestDetectTestCommandSysExecutable:
+    def test_detect_pyproject_returns_sys_executable_pytest(self, tmp_path):
+        """pyproject.toml project must get sys.executable -m pytest, not bare pytest."""
+        from agents.runner import _detect_test_command
+
+        (tmp_path / "pyproject.toml").write_text("[build-system]\n")
+        result = _detect_test_command(str(tmp_path))
+        assert result == f"{sys.executable} -m pytest"
+
+    def test_detect_setup_py_returns_sys_executable_pytest(self, tmp_path):
+        """setup.py project must get sys.executable -m pytest, not bare pytest."""
+        from agents.runner import _detect_test_command
+
+        (tmp_path / "setup.py").write_text("from setuptools import setup\nsetup()\n")
+        result = _detect_test_command(str(tmp_path))
+        assert result == f"{sys.executable} -m pytest"
+
+    def test_detect_pyproject_not_bare_pytest(self, tmp_path):
+        """Regression: _detect_test_command must not return bare 'pytest' for pyproject.toml."""
+        from agents.runner import _detect_test_command
+
+        (tmp_path / "pyproject.toml").write_text("[build-system]\n")
+        result = _detect_test_command(str(tmp_path))
+        assert result != "pytest", (
+            "Must not use bare 'pytest' — would resolve to wrong Python version"
+        )
+
+
+# ---------------------------------------------------------------------------
+# Shared fixture for pipeline auto-test tests
+# ---------------------------------------------------------------------------
+
+@pytest.fixture
+def conn_kin109():
+    c = init_db(":memory:")
+    models.create_project(c, "proj", "Proj", "/tmp/proj", tech_stack=["python"])
+    models.create_task(c, "PROJ-001", "proj", "Fix bug", brief={"route_type": "debug"})
+    yield c
+    c.close()
+
+
+def _claude_mock_success():
+    m = MagicMock()
+    m.stdout = '{"result": "done", "usage": {"total_tokens": 100}, "cost_usd": 0.001}'
+    m.stderr = ""
+    m.returncode = 0
+    return m
+
+
+# ---------------------------------------------------------------------------
+# (4/5) Pipeline auto-test: project.test_command priority vs auto-detect
+# ---------------------------------------------------------------------------
+
+class TestAutoTestCommandPriority:
+    @patch("agents.runner._run_project_tests")
+    @patch("agents.runner.subprocess.run")
+    def test_uses_project_test_command_when_set(self, mock_run, mock_tests, conn_kin109, tmp_path):
+        """When project.test_command is set, pipeline auto-test must use that command."""
+        mock_run.return_value = _claude_mock_success()
+        mock_tests.return_value = {"success": True, "output": "ok", "returncode": 0}
+
+        (tmp_path / "pyproject.toml").write_text("[build-system]\n")
+        models.update_project(
+            conn_kin109, "proj",
+            test_command="my_custom_runner",
+            auto_test_enabled=True,
+            path=str(tmp_path),
+        )
+
+        from agents.runner import run_pipeline
+        run_pipeline(conn_kin109, "PROJ-001", [{"role": "backend_dev", "brief": "fix it"}])
+
+        mock_tests.assert_called_once()
+        assert mock_tests.call_args.args[1] == "my_custom_runner"
+
+    @patch("agents.runner._run_project_tests")
+    @patch("agents.runner.subprocess.run")
+    def test_uses_detect_when_test_command_null(self, mock_run, mock_tests, conn_kin109, tmp_path):
+        """When project.test_command is NULL, pipeline auto-test uses _detect_test_command result."""
+        mock_run.return_value = _claude_mock_success()
+        mock_tests.return_value = {"success": True, "output": "ok", "returncode": 0}
+
+        (tmp_path / "pyproject.toml").write_text("[build-system]\n")
+        models.update_project(
+            conn_kin109, "proj",
+            test_command=None,
+            auto_test_enabled=True,
+            path=str(tmp_path),
+        )
+
+        from agents.runner import run_pipeline
+        run_pipeline(conn_kin109, "PROJ-001", [{"role": "backend_dev", "brief": "fix it"}])
+
+        mock_tests.assert_called_once()
+        assert mock_tests.call_args.args[1] == f"{sys.executable} -m pytest"
+
+
+# ---------------------------------------------------------------------------
+# (6) PATCH /api/projects: test_command absent → DB field unchanged (decision #580)
+# ---------------------------------------------------------------------------
+
+class TestPatchProjectTestCommand:
+    @pytest.fixture
+    def client(self, tmp_path):
+        import web.api as api_module
+        api_module.DB_PATH = tmp_path / "test.db"
+        from fastapi.testclient import TestClient
+        from web.api import app
+        c = TestClient(app)
+        c.post("/api/projects", json={"id": "px", "name": "PX", "path": "/px"})
+        # Pre-set a test_command value
+        c.patch("/api/projects/px", json={"test_command": "existing_runner"})
+        return c
+
+    def test_patch_without_test_command_does_not_change_db(self, client):
+        """PATCH without test_command field must NOT overwrite the existing value."""
+        r = client.patch("/api/projects/px", json={"auto_test_enabled": False})
+        assert r.status_code == 200
+
+        r2 = client.get("/api/projects/px")
+        assert r2.status_code == 200
+        assert r2.json()["test_command"] == "existing_runner"
+
+    def test_patch_with_test_command_updates_db(self, client):
+        """PATCH with test_command sets the new value in DB."""
+        r = client.patch("/api/projects/px", json={"test_command": "new_runner"})
+        assert r.status_code == 200
+
+        r2 = client.get("/api/projects/px")
+        assert r2.json()["test_command"] == "new_runner"
+
+    def test_patch_with_empty_test_command_stores_null(self, client):
+        """PATCH with test_command='' clears it to NULL (enables auto-detect)."""
+        r = client.patch("/api/projects/px", json={"test_command": ""})
+        assert r.status_code == 200
+
+        r2 = client.get("/api/projects/px")
+        assert r2.json()["test_command"] is None
+
+
+# ---------------------------------------------------------------------------
+# (7) Regression: auto-test early return (no framework detected) does not crash
+# ---------------------------------------------------------------------------
+
+class TestAutoTestEarlyReturn:
+    @patch("agents.runner.subprocess.run")
+    def test_pipeline_skips_auto_test_when_no_framework_detected(self, mock_run, conn_kin109, tmp_path):
+        """When no test framework is detected and test_command is NULL, pipeline must
+        skip auto-test and return success — not crash."""
+        mock_run.return_value = _claude_mock_success()
+
+        # tmp_path has no Makefile, pyproject.toml, package.json, setup.py, tsconfig.json
+        # → _detect_test_command returns None → auto-test skipped
+        models.update_project(
+            conn_kin109, "proj",
+            test_command=None,
+            auto_test_enabled=True,
+            path=str(tmp_path),
+        )
+
+        from agents.runner import run_pipeline
+        result = run_pipeline(conn_kin109, "PROJ-001", [{"role": "backend_dev", "brief": "fix"}])
+
+        assert isinstance(result, dict)
+        assert result["success"] is True
+        # The skipped auto-test step must be in results
+        skipped = [r for r in result["results"] if r.get("_skipped") and r.get("_project_test")]
+        assert len(skipped) == 1
diff --git a/tests/test_kin_110_regression.py b/tests/test_kin_110_regression.py
new file mode 100644
index 0000000..c1d9cc2
--- /dev/null
+++ b/tests/test_kin_110_regression.py
@@ -0,0 +1,162 @@
+"""Regression tests for KIN-110 — Followup button for blocked tasks.
+
+Verifies:
+1. POST /api/tasks/{id}/followup endpoint exists and responds for blocked task
+2. Endpoint returns 404 for non-existent task
+3. Endpoint works for any task status (not just blocked)
+4. Response contains expected fields: created, pending_actions, needs_decision
+"""
+
+import json
+import pytest
+from unittest.mock import patch, MagicMock
+
+import web.api as api_module
+
+
+@pytest.fixture
+def client(tmp_path):
+    db_path = tmp_path / "test.db"
+    api_module.DB_PATH = db_path
+    from web.api import app
+    from fastapi.testclient import TestClient
+    c = TestClient(app)
+    c.post("/api/projects", json={"id": "p1", "name": "P1", "path": "/p1"})
+    c.post("/api/tasks", json={"project_id": "p1", "title": "Fix auth"})
+    return c
+
+
+@pytest.fixture
+def blocked_client(tmp_path):
+    """Client with a blocked task seeded."""
+    db_path = tmp_path / "test.db"
+    api_module.DB_PATH = db_path
+    from web.api import app
+    from fastapi.testclient import TestClient
+    from core.db import init_db
+    from core import models
+    c = TestClient(app)
+    c.post("/api/projects", json={"id": "p1", "name": "P1", "path": "/p1"})
+    c.post("/api/tasks", json={"project_id": "p1", "title": "Blocked task"})
+    # Set task to blocked status
+    conn = init_db(db_path)
+    models.update_task(conn, "P1-001", status="blocked")
+    conn.close()
+    return c
+
+
+class TestFollowupEndpoint:
+    """Tests for POST /api/tasks/{id}/followup endpoint."""
+
+    @patch("agents.runner._run_claude")
+    def test_followup_blocked_task_returns_200(self, mock_claude, blocked_client):
+        """POST /followup на blocked-задаче должен вернуть 200."""
+        mock_claude.return_value = {
+            "output": json.dumps([
+                {"title": "Fix dependency", "type": "hotfix", "priority": 2,
+                 "brief": "Add missing dep"},
+            ]),
+            "returncode": 0,
+        }
+        r = blocked_client.post("/api/tasks/P1-001/followup", json={})
+        assert r.status_code == 200
+
+    @patch("agents.runner._run_claude")
+    def test_followup_returns_created_tasks(self, mock_claude, blocked_client):
+        """Ответ /followup содержит список created с созданными задачами."""
+        mock_claude.return_value = {
+            "output": json.dumps([
+                {"title": "Fix dependency", "type": "hotfix", "priority": 2,
+                 "brief": "Add missing dep"},
+            ]),
+            "returncode": 0,
+        }
+        r = blocked_client.post("/api/tasks/P1-001/followup", json={})
+        assert r.status_code == 200
+        data = r.json()
+        assert "created" in data
+        assert len(data["created"]) == 1
+        assert data["created"][0]["title"] == "Fix dependency"
+
+    @patch("agents.runner._run_claude")
+    def test_followup_response_has_required_fields(self, mock_claude, blocked_client):
+        """Ответ /followup содержит поля created, pending_actions, needs_decision."""
+        mock_claude.return_value = {"output": "[]", "returncode": 0}
+        r = blocked_client.post("/api/tasks/P1-001/followup", json={})
+        assert r.status_code == 200
+        data = r.json()
+        assert "created" in data
+        assert "pending_actions" in data
+        assert "needs_decision" in data
+
+    def test_followup_nonexistent_task_returns_404(self, blocked_client):
+        """POST /followup на несуществующую задачу → 404."""
+        r = blocked_client.post("/api/tasks/NOPE/followup", json={})
+        assert r.status_code == 404
+
+    @patch("agents.runner._run_claude")
+    def test_followup_dry_run_does_not_create_tasks(self, mock_claude, blocked_client):
+        """dry_run=true не создаёт задачи в БД."""
+        from core.db import init_db
+        from core import models
+        mock_claude.return_value = {
+            "output": json.dumps([
+                {"title": "Dry run task", "type": "hotfix", "priority": 3,
+                 "brief": "should not be created"},
+            ]),
+            "returncode": 0,
+        }
+        r = blocked_client.post("/api/tasks/P1-001/followup", json={"dry_run": True})
+        assert r.status_code == 200
+
+        # No tasks should be created in DB (only original P1-001 remains)
+        conn = init_db(api_module.DB_PATH)
+        tasks = models.list_tasks(conn, project_id="p1")
+        conn.close()
+        assert len(tasks) == 1  # Only P1-001, no new tasks
+
+    @patch("agents.runner._run_claude")
+    def test_followup_creates_child_tasks_in_db(self, mock_claude, blocked_client):
+        """Созданные followup-задачи сохраняются в БД с parent_task_id."""
+        from core.db import init_db
+        from core import models
+        mock_claude.return_value = {
+            "output": json.dumps([
+                {"title": "New dep task", "type": "feature", "priority": 4,
+                 "brief": "Install dependency"},
+            ]),
+            "returncode": 0,
+        }
+        r = blocked_client.post("/api/tasks/P1-001/followup", json={})
+        assert r.status_code == 200
+
+        conn = init_db(api_module.DB_PATH)
+        tasks = models.list_tasks(conn, project_id="p1")
+        conn.close()
+        # Original task + 1 followup
+        assert len(tasks) == 2
+        followup = next((t for t in tasks if t["id"] != "P1-001"), None)
+        assert followup is not None
+        assert followup["parent_task_id"] == "P1-001"
+
+    @patch("agents.runner._run_claude")
+    def test_followup_pending_actions_for_permission_blocked_items(
+        self, mock_claude, blocked_client
+    ):
+        """Элементы с permission-блокером попадают в pending_actions, не в created."""
+        mock_claude.return_value = {
+            "output": json.dumps([
+                {"title": "Normal task", "type": "hotfix", "priority": 2,
+                 "brief": "Just a fix"},
+                {"title": "Ручное применение .env",
+                 "type": "hotfix", "priority": 3,
+                 "brief": "Не получили разрешение на запись"},
+            ]),
+            "returncode": 0,
+        }
+        r = blocked_client.post("/api/tasks/P1-001/followup", json={})
+        assert r.status_code == 200
+        data = r.json()
+        assert len(data["created"]) == 1
+        assert len(data["pending_actions"]) == 1
+        assert data["needs_decision"] is True