fix: bind Cua pixel actions to current screenshot geometry

Correct native scroll scaling, reject stale or out-of-bounds pixel targets, and scope DPI awareness to geometry reads. Package adapter 0.36.0+artemis.3 and document the high-effort cross-device verification.

Validation: Go tests and vet, packaged MCP schema and handshake, Python checks, and user-run high-effort task completed with a passed verdict and explicit coverage limits.

plugins/cua-driver/coordinates_test.go

@@ -0,0 +1,137 @@
+package main
+
+import (
+	"errors"
+	"math"
+	"testing"
+)
+
+func fixtureFrame(w, h int) pointerFrame {
+	return pointerFrame{"current", w, h, obj{"left": -2560, "top": 120, "width": 2560, "height": 1540}}
+}
+
+func TestScrollMapsActualNativeCaptureWithoutDPIOrWindowOffset(t *testing.T) {
+	// Reproduce the task's 2558x1538 -> 1280x770 image transition. Window
+	// position and OS logical-screen dimensions must never affect this ratio.
+	f := fixtureFrame(1280, 770)
+	args := obj{"pid": 1, "window_id": 2, "capture_id": "current", "x": 900, "y": 600, "direction": "down"}
+	got, err := nativeScrollArgs(f, args, func() (pointerFrame, error) { return fixtureFrame(2558, 1538), nil })
+	if err != nil {
+		t.Fatal(err)
+	}
+	if math.Abs(got["x"].(float64)-1798.59375) > 0.0001 || math.Abs(got["y"].(float64)-1198.441558) > 0.0001 {
+		t.Fatalf("incorrect mapping: %v", got)
+	}
+	if _, ok := got["capture_id"]; ok {
+		t.Fatal("upstream scroll does not support capture_id")
+	}
+	if args["x"] != 900 || args["capture_id"] != "current" {
+		t.Fatal("input evidence mutated")
+	}
+	if got["pid"] != 1 || got["window_id"] != 2 || got["direction"] != "down" {
+		t.Fatal("target changed")
+	}
+}
+
+func TestNativeScrollAndNonCroppedCapture(t *testing.T) {
+	for _, size := range [][2]int{{2558, 1538}, {2560, 1540}, {1280, 770}} {
+		f := fixtureFrame(size[0], size[1])
+		args := obj{"capture_id": "current", "x": 200.5, "y": 300.25}
+		got, err := nativeScrollArgs(f, args, func() (pointerFrame, error) { return f, nil })
+		if err != nil || got["x"] != 200.5 || got["y"] != 300.25 {
+			t.Fatalf("native coordinates changed: %v %v", got, err)
+		}
+	}
+	// WGC can have a different native extent than a one-pixel DWM crop.
+	f := fixtureFrame(1280, 770)
+	got, err := nativeScrollArgs(f, obj{"capture_id": "current", "x": 640, "y": 385}, func() (pointerFrame, error) { return fixtureFrame(2560, 1540), nil })
+	if err != nil || got["x"] != float64(1280) || got["y"] != float64(770) {
+		t.Fatalf("assumed DWM crop: %v %v", got, err)
+	}
+}
+
+func TestInvalidCoordinatesNeverReachNativeProbe(t *testing.T) {
+	f := fixtureFrame(1280, 770)
+	for _, bad := range []obj{
+		{"capture_id": "current", "x": 1700, "y": 1000}, // Actual failing task.
+		{"capture_id": "old", "x": 900, "y": 600},
+		{"x": 900, "y": 600},
+		{"capture_id": "current", "x": -1, "y": 0},
+		{"capture_id": "current", "x": 1280, "y": 0},
+		{"capture_id": "current", "x": 0, "y": 770},
+		{"capture_id": "current", "x": "900", "y": 600},
+		{"capture_id": "current", "x": math.NaN(), "y": 600},
+		{"capture_id": "current", "x": 900, "y": math.Inf(1)},
+	} {
+		_, err := nativeScrollArgs(f, bad, func() (pointerFrame, error) { t.Fatal("invalid coordinates reached probe"); return f, nil })
+		if err == nil {
+			t.Fatalf("accepted %v", bad)
+		}
+	}
+}
+
+func TestScrollRejectsMovedResizedOrDifferentCrop(t *testing.T) {
+	f := fixtureFrame(1280, 770)
+	for _, kind := range []string{"moved", "resized", "crop", "missing"} {
+		native := fixtureFrame(2558, 1538)
+		switch kind {
+		case "moved":
+			native.window["left"] = 0
+		case "resized":
+			native.window["width"] = 1920
+		case "crop":
+			native.height = 1000
+		case "missing":
+			native.width = 0
+		}
+		_, err := nativeScrollArgs(f, obj{"capture_id": "current", "x": 900, "y": 600}, func() (pointerFrame, error) { return native, nil })
+		if err == nil {
+			t.Fatal(kind)
+		}
+	}
+	_, err := nativeScrollArgs(f, obj{"capture_id": "current", "x": 900, "y": 600}, func() (pointerFrame, error) { return pointerFrame{}, errors.New("capture disconnected") })
+	if err == nil {
+		t.Fatal("missing capture was accepted")
+	}
+}
+
+func TestClickAndBothDragEndpointsUseLatestFrame(t *testing.T) {
+	f := fixtureFrame(1280, 770)
+	click := obj{"capture_id": "current", "x": 1279, "y": 769}
+	if err := f.validate("click_at", click); err != nil {
+		t.Fatal(err)
+	}
+	if click["x"] != 1279 || click["y"] != 769 {
+		t.Fatal("click scaled twice")
+	}
+	drag := obj{"capture_id": "current", "from_x": 0, "from_y": 0, "to_x": 1279, "to_y": 769}
+	if err := f.validate("drag", drag); err != nil {
+		t.Fatal(err)
+	}
+	for _, key := range []string{"from_x", "from_y", "to_x", "to_y"} {
+		bad := copyArgs(drag)
+		bad[key] = 3000
+		if f.validate("drag", bad) == nil {
+			t.Fatal("accepted out of bounds " + key)
+		}
+	}
+	drag["capture_id"] = "old"
+	if f.validate("drag", drag) == nil {
+		t.Fatal("accepted stale drag")
+	}
+}
+
+func TestAllPixelSchemasRequireCapture(t *testing.T) {
+	a := newAdapter(nil)
+	for _, name := range []string{"click_at", "scroll", "drag"} {
+		found := false
+		for _, key := range items(a.schemas[name]["required"]) {
+			if key == "capture_id" {
+				found = true
+			}
+		}
+		if !found {
+			t.Fatal(name + " lacks capture binding")
+		}
+	}
+}