fix(cli): send an [x,y,z] --view-target to Unity as a position

manage_camera reads viewTarget as a GameObject reference, or as a
position when it is a JSON array. The CLI sent the option as text, so
"[0, 1, 2]" reached Unity as a GameObject name, although the help offered
[x,y,z]. A value that starts with "[" is now parsed as a JSON array, for
both screenshot and screenshot-multiview, and the guide no longer says a
position needs `raw`.

Author: Scriptwonder

Server/src/cli/commands/camera.py

@@ -6,7 +6,7 @@
 from cli.utils.config import get_config
 from cli.utils.output import format_output, print_error, print_success
 from cli.utils.connection import run_command, handle_unity_errors
-from cli.utils.parsers import parse_json_dict_or_exit
+from cli.utils.parsers import parse_json_dict_or_exit, parse_json_list_or_exit
 from cli.utils.constants import SEARCH_METHOD_CHOICE_BASIC
 
 
@@ -519,8 +519,11 @@ def screenshot(camera_ref, file_name, super_size, include_image, max_resolution,
         params["captureSource"] = capture_source
     if batch:
         params["batch"] = batch
+    # Unity reads a JSON array as a position and any other value as a GameObject; the
+    # option arrives as text, so "[x,y,z]" is parsed to reach Unity as an array.
     if view_target:
-        params["viewTarget"] = view_target
+        params["viewTarget"] = (parse_json_list_or_exit(view_target, "view-target")
+                                if view_target.lstrip().startswith("[") else view_target)
     if output_folder:
         params["outputFolder"] = output_folder
     result = run_command("manage_camera", params, config)
@@ -529,7 +532,7 @@ def screenshot(camera_ref, file_name, super_size, include_image, max_resolution,
 
 @camera.command("screenshot-multiview")
 @click.option("--max-resolution", type=int, default=None, help="Max resolution per tile.")
-@click.option("--view-target", default=None, help="Center target for the multiview capture.")
+@click.option("--view-target", default=None, help="Center target: GameObject name/path/ID or [x,y,z].")
 @click.option("--output-folder", default=None,
               help="Output folder, project-relative or absolute inside the project. "
                    "Overrides Editor preference; falls back to Assets/Screenshots when unset.")
@@ -547,7 +550,8 @@ def screenshot_multiview(max_resolution, view_target, output_folder):
     if max_resolution is not None:
         params["maxResolution"] = max_resolution
     if view_target:
-        params["viewTarget"] = view_target
+        params["viewTarget"] = (parse_json_list_or_exit(view_target, "view-target")
+                                if view_target.lstrip().startswith("[") else view_target)
     if output_folder:
         params["outputFolder"] = output_folder
     result = run_command("manage_camera", params, config)

Server/tests/test_cli.py

@@ -476,6 +476,13 @@ def test_camera_screenshot_scene_view(self, runner, mock_unity_response):
             assert params["viewTarget"] == "Canvas"
             assert params["includeImage"] is True
 
+    def test_camera_screenshot_position_view_target_reaches_unity_as_an_array(self, runner, mock_unity_response):
+        # Unity reads a JSON array as a position; the text "[0, 1, 2]" would be looked up as a GameObject name.
+        with patch("cli.commands.camera.run_command", return_value=mock_unity_response) as mock_run:
+            result = runner.invoke(cli, ["camera", "screenshot", "--view-target", "[0, 1, 2]"])
+            assert result.exit_code == 0, result.output
+            assert mock_run.call_args[0][1]["viewTarget"] == [0, 1, 2]
+
     def test_camera_ping_prints_output(self, runner, mock_unity_response):
         """The camera group must actually emit its result.
 

website/docs/guides/cli.md

@@ -392,10 +392,10 @@ unity-mcp camera screenshot-multiview --view-target "Player" --max-resolution 48
 | `--max-resolution` | int | Longest edge of the inline image in pixels (default 640; 480 per tile with `--batch`) |
 | `--capture-source` | string | `game_view` (default) or `scene_view` (editor viewport) |
 | `--batch` | string | `surround` (6 angles) or `orbit` (grid around the target) |
-| `--view-target` | string | GameObject name/path/ID. Aims the camera (game_view) or frames the Scene View (scene_view) |
+| `--view-target` | string | GameObject name/path/ID, or a `[x,y,z]` position. Aims the camera (game_view) or frames the Scene View (scene_view) |
 | `--output-folder` | string | Save folder, project-relative or absolute inside the project (default: Editor preference, then `Assets/Screenshots`) |
 
-`camera screenshot-multiview` takes `--max-resolution`, `--view-target` and `--output-folder`. `--view-target` is sent as a string, so it cannot carry a `[x,y,z]` position; that, and the MCP tool's `view_position`, `view_rotation` and `orbit_*` settings, need `unity-mcp raw manage_camera '{"action": "screenshot", ...}'`.
+`camera screenshot-multiview` takes `--max-resolution`, `--view-target` and `--output-folder`. The MCP tool's `view_position`, `view_rotation` and `orbit_*` settings need `unity-mcp raw manage_camera '{"action": "screenshot", ...}'`.
 
 ### Graphics Operations