feat(mcp): write, test and delete this instance's own playbooks (#5748)

#5748 "MCP can write, edit and test this instance's own playbooks
(steward-local)": two admin tools beside run_playbook, so an agent can
build automation for these hosts that isn't part of Steward itself.

- save_playbook(playbook, content, source="steward-local", replace=false)
  goes through the editor's own validate_playbook_yaml and
  write_playbook. A playbook that isn't a list of plays comes back with
  the YAML error and its line, and nothing is written. An existing file
  is overwritten only with replace=true. steward-builtin is refused
  (Steward's playbooks change in its repository), as are git sources
  (their repo owns them).
- After writing, `ansible-playbook --syntax-check` runs against a
  throwaway localhost inventory (no host contacted, 60 s), so a module
  name Ansible can't resolve shows before any run. The result rides the
  reply; the file is kept either way.
- delete_playbook(playbook, source="steward-local").
- Both audit as ansible.playbook.created/updated/deleted, "<user> via
  MCP", with the size and never the content.

The testing loop is the existing one: run_playbook defaults to check
mode, on one target, then get_ansible_run. The integration test does
that loop with the real ansible-playbook.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/integration/test_mcp_automation.py

@@ -197,3 +197,72 @@ def test_runs_schedules_and_alert_actions_through_mcp(app, tmp_path):
         assert len(await _audit("alert_rule.action_cleared", rule_name)) == 1
 
     asyncio.run(_go())
+
+
+@_NEEDS_DB
+def test_an_instance_playbook_is_written_checked_run_and_removed_through_mcp(app, tmp_path, monkeypatch):
+    """#5748: save_playbook into steward-local, the real syntax check, a
+    check-mode run of what was saved, replace, delete; audit rows carry the
+    size and never the content."""
+    from sqlalchemy import select
+    from steward.ansible import sources as src_module
+    from steward.core.api_tokens import Caller
+    from steward.mcp import _context
+    from steward.mcp.tools import automation as a
+    from steward.models.ansible_inventory import AnsibleTarget
+    from steward.models.audit import AuditEvent
+    from steward.models.users import User, UserRole
+
+    monkeypatch.setattr(src_module, "USER_PLAYBOOK_DIR", str(tmp_path / "local"))
+    tag = uuid.uuid4().hex[:8]
+    uid, username, target_name = str(uuid.uuid4()), f"mcp-pb-{tag}", f"mcp-pb-target-{tag}"
+    marker = "MARKER-" + tag
+    good = textwrap.dedent(f"""\
+        - hosts: all
+          connection: local
+          gather_facts: false
+          tasks:
+            - ansible.builtin.debug:
+                msg: "{marker}"
+    """)
+    typo = good.replace("ansible.builtin.debug", "ansible.builtin.debugg")
+
+    async def _go():
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                s.add(User(id=uid, username=username, email=f"{tag}@e.test",
+                           password_hash="x", role=UserRole.admin, is_active=True))
+                s.add(AnsibleTarget(name=target_name, address="127.0.0.1",
+                                    ansible_vars={"ansible_connection": "local"}))
+        _context._app_ctx.set(app)
+        _context._caller_ctx.set(Caller(user_id=uid, username=username,
+                                        role=UserRole.admin, token_id=7))
+
+        bad = await a.save_playbook("upkeep/check.yml", typo)
+        assert bad["syntax"]["ok"] is False and "debugg" in bad["syntax"]["output"]
+        fixed = await a.save_playbook("upkeep/check.yml", good, replace=True)
+        assert fixed["syntax"] == {"ok": True} and fixed["replaced"] is True
+
+        started = await a.run_playbook("steward-local", "upkeep/check.yml",
+                                       scope=f"target:{target_name}")
+        assert started["check"] is True
+        done = await a.wait_for_run(started["run_id"], timeout_s=120)
+        assert done["status"] == "success", done["tail"]
+        assert marker in "\n".join(done["tail"])
+
+        assert await a.delete_playbook("upkeep/check.yml") == {
+            "deleted": "upkeep/check.yml", "source": "steward-local"}
+        assert "upkeep/check.yml" not in [
+            b["playbook"] for b in (await a.list_playbooks(source="steward-local"))["playbooks"]]
+
+        async with app.db_sessionmaker() as s:
+            rows = list((await s.execute(select(AuditEvent).where(
+                AuditEvent.user_id == uid, AuditEvent.action.like("ansible.playbook.%"))
+                .order_by(AuditEvent.timestamp))).scalars())
+        assert [r.action for r in rows] == ["ansible.playbook.created", "ansible.playbook.updated",
+                                           "ansible.playbook.deleted"]
+        assert all(r.username == f"{username} via MCP" for r in rows)
+        assert json.loads(rows[1].detail_json)["bytes"] == len(good)
+        assert not any(marker in (r.detail_json or "") for r in rows)
+
+    asyncio.run(_go())

tests/mcp/test_automation.py

@@ -25,6 +25,7 @@ VIEWER_TOOLS = {"list_playbooks", "get_playbook", "list_inventory", "list_ansibl
                 "get_ansible_run", "list_schedules", "wait_for_run"}
 OPERATOR_TOOLS = {"run_playbook", "cancel_ansible_run", "create_schedule", "update_schedule",
                   "delete_schedule", "update_group_members"}
+ADMIN_TOOLS = {"set_alert_action", "save_playbook", "delete_playbook"}
 
 
 def test_split_vars_keeps_secrets_apart_from_what_is_stored():
@@ -96,8 +97,10 @@ def test_automation_tools_are_registered_with_their_roles(app):
     for name in OPERATOR_TOOLS:
         assert mcp.tool_roles[name] is UserRole.operator, name
         assert mcp._tool_manager.get_tool(name).annotations.read_only_hint is False, name
-    assert mcp.tool_roles["set_alert_action"] is UserRole.admin
-    for name in VIEWER_TOOLS | OPERATOR_TOOLS | {"set_alert_action"}:
+    for name in ADMIN_TOOLS:
+        assert mcp.tool_roles[name] is UserRole.admin, name
+        assert mcp._tool_manager.get_tool(name).annotations.read_only_hint is False, name
+    for name in VIEWER_TOOLS | OPERATOR_TOOLS | ADMIN_TOOLS:
         assert (mcp._tool_manager.get_tool(name).description or "").strip(), name
 
 
@@ -107,6 +110,9 @@ def test_a_viewer_token_cannot_start_a_run(app):
     assert "needs the operator role" in app.mcp_server.refusal("run_playbook", viewer)
     assert app.mcp_server.refusal("run_playbook", operator) is None
     assert "needs the admin role" in app.mcp_server.refusal("set_alert_action", operator)
+    for name in ("save_playbook", "delete_playbook"):
+        assert "needs the admin role" in app.mcp_server.refusal(name, operator)
+        assert "needs the admin role" in app.mcp_server.refusal(name, viewer)
 
 
 def test_builtin_playbooks_get_steward_url_from_steward():
@@ -168,3 +174,97 @@ def test_group_members_are_named_by_target_or_by_linked_host():
     assert [t.name for t in got] == ["web01", "db01"]
     with pytest.raises(ValueError, match="no target or linked host named 'nas'.*Nothing was changed"):
         automation.member_targets(["web01", "nas"], [web, db], hosts)
+
+
+# ── writing this instance's own playbooks (#5748) ───────────────────────────
+
+PLAY = "- hosts: all\n  gather_facts: false\n  tasks:\n    - ansible.builtin.ping:\n"
+
+
+@pytest.fixture
+def writable(monkeypatch, tmp_path):
+    """steward-local in a temp folder beside the read-only sources, with the
+    audit rows and the syntax check captured."""
+    from steward.ansible import sources as src_module
+    local = tmp_path / "local"
+    srcs = [{"name": src_module.BUILTIN_SOURCE_NAME, "type": "local", "path": str(tmp_path / "builtin")},
+            {"name": src_module.LOCAL_SOURCE_NAME, "type": "local", "path": str(local)},
+            {"name": "infra", "type": "git", "path": str(tmp_path / "git")}]
+    monkeypatch.setattr(automation, "_sources", lambda: srcs)
+    seen = {"audits": [], "checked": []}
+
+    async def audit(action, entity_type=None, entity_id=None, detail=None):
+        seen["audits"].append((action, entity_id, detail))
+
+    async def syntax(path, rel):
+        seen["checked"].append(rel)
+        return seen.get("syntax", (True, ""))
+    monkeypatch.setattr(automation, "audit_mcp", audit)
+    monkeypatch.setattr(automation.executor, "syntax_check", syntax)
+    seen["dir"] = local
+    return seen
+
+
+async def test_a_new_playbook_is_written_checked_and_audited_without_its_content(writable):
+    out = await automation.save_playbook("upkeep/ping.yml", PLAY)
+    assert (writable["dir"] / "upkeep/ping.yml").read_text() == PLAY
+    assert out["source"] == "steward-local" and out["replaced"] is False
+    assert out["syntax"] == {"ok": True} and writable["checked"] == ["upkeep/ping.yml"]
+    [(action, entity, detail)] = writable["audits"]
+    assert action == "ansible.playbook.created" and entity == "steward-local:upkeep/ping.yml"
+    assert detail == {"source": "steward-local", "playbook": "upkeep/ping.yml", "bytes": len(PLAY)}
+    assert PLAY not in str(writable["audits"])
+
+
+async def test_an_existing_playbook_is_replaced_only_when_asked(writable):
+    await automation.save_playbook("ping.yml", PLAY)
+    with pytest.raises(ValueError, match="already has 'ping.yml'.*replace=true.*Nothing was saved"):
+        await automation.save_playbook("ping.yml", "- hosts: web\n  tasks: []\n")
+    assert (writable["dir"] / "ping.yml").read_text() == PLAY
+    out = await automation.save_playbook("ping.yml", "- hosts: web\n  tasks: []\n", replace=True)
+    assert out["replaced"] is True
+    assert [a for a, _, _ in writable["audits"]] == ["ansible.playbook.created", "ansible.playbook.updated"]
+
+
+@pytest.mark.parametrize("content, why", [
+    ("- hosts: all\n  tasks: [unclosed\n", "(?s)YAML error.*line \\d"),
+    ("hosts: all\n", "list of plays"),
+    ("", "empty"),
+])
+async def test_a_playbook_that_isnt_one_says_why_and_writes_nothing(writable, content, why):
+    with pytest.raises(ValueError, match=why):
+        await automation.save_playbook("bad.yml", content)
+    assert not (writable["dir"] / "bad.yml").exists()
+    assert writable["audits"] == [] and writable["checked"] == []
+
+
+async def test_stewards_own_and_git_sources_are_refused(writable):
+    with pytest.raises(ValueError, match="Steward's own playbooks.*repository.*steward-local"):
+        await automation.save_playbook("x.yml", PLAY, source="steward-builtin")
+    with pytest.raises(ValueError, match="git repository"):
+        await automation.save_playbook("x.yml", PLAY, source="infra")
+    with pytest.raises(ValueError, match="Steward's own playbooks"):
+        await automation.delete_playbook("x.yml", source="steward-builtin")
+    for bad in ("../escape.yml", "notes.txt"):
+        with pytest.raises(ValueError, match="Nothing was saved"):
+            await automation.save_playbook(bad, PLAY)
+    assert writable["audits"] == []
+
+
+async def test_a_failed_syntax_check_is_reported_and_the_file_kept(writable):
+    writable["syntax"] = (False, "ERROR! couldn't resolve module/action 'ansible.builtin.pingg'")
+    out = await automation.save_playbook("ping.yml", PLAY)
+    assert out["syntax"] == {"ok": False, "output": writable["syntax"][1]}
+    assert (writable["dir"] / "ping.yml").exists()
+    writable["syntax"] = (None, "ansible-playbook isn't installed here, so only the YAML was checked.")
+    out = await automation.save_playbook("ping.yml", PLAY, replace=True)
+    assert out["syntax"] == {"ok": None, "note": writable["syntax"][1]}
+
+
+async def test_a_playbook_is_deleted_and_audited(writable):
+    await automation.save_playbook("ping.yml", PLAY)
+    assert await automation.delete_playbook("ping.yml") == {"deleted": "ping.yml", "source": "steward-local"}
+    assert not (writable["dir"] / "ping.yml").exists()
+    assert writable["audits"][-1][:2] == ("ansible.playbook.deleted", "steward-local:ping.yml")
+    with pytest.raises(ValueError, match="has no playbook 'ping.yml'"):
+        await automation.delete_playbook("ping.yml")