feat(catalog): cover Windows guest-mutating tools in the starter list

The catalog is opt-in: an unlisted tool passes through unguarded. The
shipped starter list was ssh/exec and ansible/run_playbook, both Linux
shaped, so an agent changing a protected Windows guest over WinRM or
PowerShell was never guarded at all. That does not fail loudly, it
simply never inserts a checkpoint.

Adds winrm/run_command, winrm/run_ps, powershell/invoke_command and
smb/write_file. The smb entry is there because a file written into a
share changes the guest without any shell being involved.

Test asserts the example config covers both platforms and that every
entry names a vm_arg, since without one the guard cannot resolve a VM.

Still illustrative, not exhaustive: tool names vary per MCP server, so
users add their own with zerto_add_mutating_tool. That reactive model is
the real weakness here and is worth revisiting separately.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_016yVfC5nvZowoLFnEGWhLGn

README.md

@@ -9,7 +9,7 @@ If the loop works, these tools are the delta to add to official ZVM MCP (`ZVM.MC
 1. `zerto_find_protection` — VM name, hostname, or vmIdentifier to exactly one VM and every VPG. Zero or two-plus VMs: stop.
 2. `zerto_create_tagged_checkpoint` / `zerto_guard_before_mutate` — same tag on every protecting VPG, wait until listed. The name records which agent and what it is doing: `ai:<agent> | <action> | vm=<vm> | change=<id> | <utc>`.
 3. `zerto_recover_file` — FLR after a human sets `confirmed=true`. Linux and Windows guest paths. Locally replicated VPGs only: FLR runs at the VPG's recovery site. Reports its own unmount; `zerto_list_flr_sessions` / `zerto_end_flr_session` find and reap a mount orphaned by a crashed recovery.
-4. Mutating catalog — opt-in list of MCP tools that must be guarded. Unlisted tools pass through. Users add entries.
+4. Mutating catalog — opt-in list of MCP tools that must be guarded. Unlisted tools pass through. Users add entries. The starter list covers Linux (`ssh`, `ansible`) and Windows (`winrm`, `powershell`, `smb`), and is illustrative, not exhaustive.
 
 Official ZVM MCP already has inventory and failover test. It does not insert tagged checkpoints or run FLR.
 

config.example.json

@@ -17,6 +17,30 @@
       "tool": "run_playbook",
       "vm_arg": "limit",
       "notes": "Playbook target host/group. Resolve to a single VM before guard."
+    },
+    {
+      "server": "winrm",
+      "tool": "run_command",
+      "vm_arg": "host",
+      "notes": "Windows remote shell over WinRM. vm_arg is the hostname."
+    },
+    {
+      "server": "winrm",
+      "tool": "run_ps",
+      "vm_arg": "host",
+      "notes": "PowerShell over WinRM. Same blast radius as run_command."
+    },
+    {
+      "server": "powershell",
+      "tool": "invoke_command",
+      "vm_arg": "computer_name",
+      "notes": "Invoke-Command against a remote Windows guest."
+    },
+    {
+      "server": "smb",
+      "tool": "write_file",
+      "vm_arg": "host",
+      "notes": "Writes a file into a Windows share. Changes the guest without a shell."
     }
   ]
 }

tests/test_catalog.py

@@ -31,3 +31,22 @@ def test_entry_requires_fields():
         raise AssertionError("expected ValueError")
     except ValueError:
         pass
+
+
+def test_example_config_covers_windows_and_linux():
+    """The starter catalog must not be Linux-only.
+
+    The catalog is opt-in: an unlisted tool passes through unguarded. A
+    Windows-only shop taking the shipped defaults would therefore mutate
+    protected guests with no checkpoint at all.
+    """
+    example = Path(__file__).resolve().parent.parent / "config.example.json"
+    data = json.loads(example.read_text(encoding="utf-8"))
+    cat = MutatingCatalog.from_config(data)
+    assert cat.get("ssh", "exec") is not None
+    assert cat.get("winrm", "run_command") is not None
+    assert cat.get("winrm", "run_ps") is not None
+    assert cat.get("powershell", "invoke_command") is not None
+    # every entry must name the arg holding the VM, or the guard cannot resolve one
+    for entry in cat.list():
+        assert entry.vm_arg, f"{entry.server}/{entry.tool} has no vm_arg"