0agent 1.0.70 → 1.0.72

package/dist/daemon.mjs

@@ -3303,24 +3303,57 @@ print("Typed successfully")
             const pyKeys = JSON.stringify(parts);
             if (targetApp && platform2() === "darwin") {
               const safeApp = targetApp.replace(/'/g, "\\'");
-              const asKey = parts[parts.length - 1] ?? "";
-              const modifiers = parts.slice(0, -1).map((k) => {
+              const mainKey = parts[parts.length - 1] ?? "";
+              const modParts = parts.slice(0, -1);
+              const KEY_CODES = {
+                down: 125,
+                up: 126,
+                left: 123,
+                right: 124,
+                enter: 36,
+                return: 36,
+                escape: 53,
+                esc: 53,
+                tab: 48,
+                delete: 51,
+                backspace: 51,
+                "delete-forward": 117,
+                space: 49,
+                home: 115,
+                end: 119,
+                pageup: 116,
+                pagedown: 121,
+                f1: 122,
+                f2: 120,
+                f3: 99,
+                f4: 118,
+                f5: 96,
+                f6: 97,
+                f7: 98,
+                f8: 100,
+                f9: 101,
+                f10: 109,
+                f11: 103,
+                f12: 111
+              };
+              const keyCode = KEY_CODES[mainKey];
+              const modifiers = modParts.map((k) => {
                 if (k === "command") return "command down";
                 if (k === "ctrl") return "control down";
                 if (k === "shift") return "shift down";
                 if (k === "option") return "option down";
                 return "";
               }).filter(Boolean).join(", ");
-              const asModStr = modifiers ? ` using {${modifiers}}` : "";
+              const usingClause = modifiers ? ` using {${modifiers}}` : "";
+              const keyStatement = keyCode !== void 0 ? `key code ${keyCode}${usingClause}` : `keystroke "${mainKey}"${usingClause}`;
               return header + `
 import subprocess, time
-# Focus target app first
+# Activate app to give it OS focus, then send key to frontmost (whatever has focus)
 subprocess.run(['osascript', '-e', 'tell application "${safeApp}" to activate'], capture_output=True)
-time.sleep(0.3)
-# Send keystroke via AppleScript (reliable \u2014 goes to the focused app, not Terminal)
-as_script = '''tell application "System Events"
-  keystroke "${asKey}"${asModStr}
-end tell'''
+time.sleep(0.5)
+as_script = """tell application "System Events"
+  ${keyStatement}
+end tell"""
 r = subprocess.run(['osascript', '-e', as_script], capture_output=True, text=True)
 if r.returncode == 0:
     print(f"Sent ${parts.join("+")} to ${safeApp}")
@@ -4759,16 +4792,23 @@ content = element.text if element else page.get_all_text()` : `content = page.ge
             `    regardless of OS keyboard focus. ALWAYS use this for native app text input.`,
             `  accessibility_click {app:"WhatsApp", element:"Send"} \u2014 click button via Accessibility API.`,
             `  hotkey {keys:"cmd+f", app:"WhatsApp"} \u2014 send hotkey to specific app (not Terminal).`,
-            `WHATSAPP WORKFLOW (use this exact sequence):`,
+            `NATIVE APP SEARCH NAVIGATION (generalised \u2014 works for WhatsApp, Spotlight, file dialogs, any search):`,
+            `  After typing in a search box, results appear in a dropdown/list.`,
+            `  Navigate them with: hotkey {keys:"down", app:"AppName"} to move to first result,`,
+            `  then hotkey {keys:"enter", app:"AppName"} to select/open it.`,
+            `  If first down+enter doesn't work, try: down, down, enter (some apps skip a header row).`,
+            `  Use wait {seconds:1} after typing to give results time to load before navigating.`,
+            `WHATSAPP EXACT WORKFLOW:`,
             `  1. open_app {app:"WhatsApp"}`,
-            `  2. hotkey {keys:"cmd+f", app:"WhatsApp"} \u2014 open search`,
-            `  3. app_type {app:"WhatsApp", text:"ContactName"} \u2014 search for contact`,
-            `  4. hotkey {keys:"enter", app:"WhatsApp"} \u2014 open the conversation`,
-            `  5. app_type {app:"WhatsApp", text:"your message"} \u2014 type message`,
-            `  6. hotkey {keys:"enter", app:"WhatsApp"} \u2014 send`,
-            `  7. accessibility_click {app:"WhatsApp", element:"Send"} \u2014 if enter doesn't send`,
+            `  2. hotkey {keys:"cmd+f", app:"WhatsApp"} \u2014 open search bar`,
+            `  3. app_type {app:"WhatsApp", text:"ContactName"} \u2014 type contact name`,
+            `  4. wait {seconds:1} \u2014 wait for search results to load`,
+            `  5. hotkey {keys:"down", app:"WhatsApp"} \u2014 move focus to first result`,
+            `  6. hotkey {keys:"enter", app:"WhatsApp"} \u2014 open the conversation`,
+            `  7. app_type {app:"WhatsApp", text:"your message"} \u2014 type message`,
+            `  8. hotkey {keys:"enter", app:"WhatsApp"} \u2014 send`,
             `NEVER use bare 'type' action for native apps \u2014 it goes to Terminal not the app.`,
-            `ALWAYS verify: browser_state after web nav, get_media_state after play/pause, read_element for page content.`
+            `ALWAYS verify: browser_state after web nav, get_media_state after play/pause.`
           );
         }
         if (isSelfMod && this.agentRoot) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "0agent",
-  "version": "1.0.70",
+  "version": "1.0.72",
   "description": "A persistent, learning AI agent that runs on your machine. An agent that learns.",
   "private": false,
   "license": "Apache-2.0",