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/tools.json

@@ -270,7 +270,7 @@
   },
   {
     "name": "drag",
-    "description": "Drag between screenshot-local coordinates. Returns actual delivery effect and fresh post-action observation.",
+    "description": "Drag between screenshot-local coordinates. Returns actual delivery effect and fresh post-action observation. Requires the latest capture_id. Stale, moved/resized or out-of-image targets are rejected before input.",
     "inputSchema": {
       "additionalProperties": false,
       "properties": {
@@ -311,6 +311,11 @@
         "window_id": {
           "description": "Exact HWND returned for this application by list_windows. Required and must belong to pid.",
           "type": "integer"
+        },
+        "capture_id": {
+          "type": "string",
+          "minLength": 1,
+          "description": "capture_id from the latest returned screenshot for this exact window. Each action returns a new screenshot; use its new ID and dimensions."
         }
       },
       "required": [
@@ -319,7 +324,8 @@
         "from_x",
         "from_y",
         "to_x",
-        "to_y"
+        "to_y",
+        "capture_id"
       ],
       "type": "object"
     },
@@ -460,7 +466,7 @@
   },
   {
     "name": "scroll",
-    "description": "Scroll at x,y in the latest window screenshot. No element_token. Use a point inside the intended scroll surface, amount=1..50, by=line/page. Chromium/Brave requires foreground. Verify actual content movement in the returned observation.",
+    "description": "Scroll at x,y in the latest window screenshot. No element_token. Use a point inside the intended scroll surface, amount=1..50, by=line/page. Chromium/Brave requires foreground. Verify actual content movement in the returned observation. Requires the latest capture_id. Stale, moved/resized or out-of-image targets are rejected before input.",
     "inputSchema": {
       "additionalProperties": false,
       "properties": {
@@ -511,6 +517,11 @@
         "y": {
           "description": "Y inside the scroll area in latest window-local screenshot pixels; never screen coordinates. Must be paired with the other coordinate.",
           "type": "number"
+        },
+        "capture_id": {
+          "type": "string",
+          "minLength": 1,
+          "description": "capture_id from the latest returned screenshot for this exact window. Each action returns a new screenshot; use its new ID and dimensions."
         }
       },
       "required": [
@@ -518,7 +529,8 @@
         "window_id",
         "direction",
         "x",
-        "y"
+        "y",
+        "capture_id"
       ],
       "type": "object"
     },
@@ -561,7 +573,7 @@
   },
   {
     "name": "click_at",
-    "description": "Click a visual target at x,y in the latest window screenshot, with its exact capture_id. Never pass an element token. Returns effect and fresh observation.",
+    "description": "Click a visual target at x,y in the latest window screenshot, with its exact capture_id. Never pass an element token. Returns effect and fresh observation. Use the latest returned image dimensions; no screen coordinates or Android normalization.",
     "inputSchema": {
       "additionalProperties": false,
       "properties": {