visionclaw 0.1.195-beta.0 → 0.1.195-dev.feat-e2e-test-system.1

package/dist/builtin-skills/catalog/phone-adb-automation/phone_input.sh

@@ -1,132 +0,0 @@
-#!/usr/bin/env bash
-# phone_input.sh - Input text on Android phone via ADB
-#
-# Usage:
-#   bash phone_input.sh <text> [options]
-#
-# Options:
-#   -d DEVICE_ID    Target specific device
-#   -m METHOD       Input method: "direct" (default), "adbkbd", "clipboard"
-#   -e              Send Enter key after input
-#   -c              Clear existing text first (select all + delete)
-#   -q              Quiet mode
-#
-# Methods:
-#   direct   - adb shell input text (English/numbers only)
-#   adbkbd   - ADBKeyboard broadcast (supports Chinese, requires ADBKeyboard app)
-#   clipboard - Set clipboard + paste (requires Android 10+, may not work on all devices)
-#
-# Notes:
-#   - Direct method does NOT support Chinese/CJK characters
-#   - Special characters (spaces, quotes) are escaped automatically
-#   - For Chinese input, install ADBKeyboard and use -m adbkbd
-#
-# Examples:
-#   bash phone_input.sh "hello world"               # English text
-#   bash phone_input.sh "hello" -e                   # Type and press Enter
-#   bash phone_input.sh "你好" -m adbkbd             # Chinese via ADBKeyboard
-#   bash phone_input.sh "new text" -c                # Clear field first, then type
-
-set -euo pipefail
-
-DEVICE=""
-METHOD="direct"
-ENTER=false
-CLEAR=false
-QUIET=false
-
-# Parse args
-POSITIONAL=()
-ARGV=("$@")
-i=0
-while [ $i -lt ${#ARGV[@]} ]; do
-    case "${ARGV[$i]}" in
-        -d) DEVICE="${ARGV[$((i+1))]}"; i=$((i+2)) ;;
-        -m) METHOD="${ARGV[$((i+1))]}"; i=$((i+2)) ;;
-        -e) ENTER=true; i=$((i+1)) ;;
-        -c) CLEAR=true; i=$((i+1)) ;;
-        -q) QUIET=true; i=$((i+1)) ;;
-        -*) echo "Unknown option: ${ARGV[$i]}" >&2; exit 1 ;;
-        *) POSITIONAL+=("${ARGV[$i]}"); i=$((i+1)) ;;
-    esac
-done
-
-if [ ${#POSITIONAL[@]} -eq 0 ]; then
-    echo "Usage: $0 <text> [-d device] [-m method] [-e] [-c] [-q]" >&2
-    exit 1
-fi
-
-TEXT="${POSITIONAL[0]}"
-
-ADB_CMD="adb"
-if [ -n "$DEVICE" ]; then
-    ADB_CMD="adb -s $DEVICE"
-fi
-
-info()  { $QUIET || echo "[phone_input] $*" >&2; }
-warn()  { echo "[phone_input] WARNING: $*" >&2; }
-error() { echo "[phone_input] ERROR: $*" >&2; }
-
-# Clear existing text if requested
-if $CLEAR; then
-    info "Clearing existing text..."
-    # Select all (Ctrl+A) then delete
-    $ADB_CMD shell input keyevent KEYCODE_MOVE_HOME
-    $ADB_CMD shell input keyevent --longpress KEYCODE_SHIFT_LEFT KEYCODE_MOVE_END 2>/dev/null || true
-    $ADB_CMD shell input keyevent KEYCODE_DEL
-    sleep 0.3
-fi
-
-case "$METHOD" in
-    direct)
-        # Escape special characters for adb shell input text
-        # Replace spaces with %s, & with \&, etc.
-        ESCAPED=$(echo "$TEXT" | sed 's/ /%s/g; s/&/\\&/g; s/</\\</g; s/>/\\>/g; s/(/\\(/g; s/)/\\)/g; s/|/\\|/g; s/;/\\;/g; s/"/\\"/g; s/'"'"'/\\'"'"'/g')
-        info "Typing (direct): $TEXT"
-        $ADB_CMD shell input text "$ESCAPED"
-        ;;
-    adbkbd)
-        info "Typing (ADBKeyboard): $TEXT"
-        # Check if ADBKeyboard is available; auto-install if missing
-        if ! $ADB_CMD shell ime list -s 2>/dev/null | grep -q "adbkeyboard"; then
-            warn "ADBKeyboard not found. Attempting auto-install..."
-            SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
-            if [ -f "$SCRIPT_DIR/phone_setup.sh" ]; then
-                bash "$SCRIPT_DIR/phone_setup.sh" ${DEVICE:+-d "$DEVICE"} -q || {
-                    error "Auto-install failed. Run manually: bash phone_setup.sh"
-                    exit 1
-                }
-                info "ADBKeyboard auto-installed successfully"
-            else
-                echo "ERROR: ADBKeyboard not installed. Run phone_setup.sh first:" >&2
-                echo "  bash phone_setup.sh" >&2
-                exit 1
-            fi
-        fi
-        $ADB_CMD shell am broadcast -a ADB_INPUT_TEXT --es msg "$TEXT" 2>/dev/null
-        ;;
-    clipboard)
-        info "Typing (clipboard): $TEXT"
-        # Try to set clipboard and paste
-        $ADB_CMD shell "echo '$TEXT' | cmd clipboard set_text" 2>/dev/null || {
-            echo "ERROR: Clipboard method not supported on this device" >&2
-            exit 1
-        }
-        # Paste: Ctrl+V
-        $ADB_CMD shell input keyevent 279  # KEYCODE_PASTE
-        ;;
-    *)
-        echo "ERROR: Unknown method '$METHOD'. Use: direct, adbkbd, clipboard" >&2
-        exit 1
-        ;;
-esac
-
-# Press Enter if requested
-if $ENTER; then
-    sleep 0.2
-    $ADB_CMD shell input keyevent KEYCODE_ENTER
-    info "Pressed Enter"
-fi
-
-info "Done."
-echo "input_ok"

package/dist/builtin-skills/catalog/phone-adb-automation/phone_launch.sh

@@ -1,166 +0,0 @@
-#!/usr/bin/env bash
-# phone_launch.sh - Launch an Android app by name alias or package name
-#
-# Usage:
-#   bash phone_launch.sh <app_name_or_package> [options]
-#
-# Options:
-#   -d DEVICE_ID    Target specific device
-#   -l              List all known app aliases
-#   -q              Quiet mode
-#
-# Supports common Chinese app aliases (case-insensitive):
-#   wechat, wecom, meituan, ctrip, xiaohongshu/xhs, bilibili,
-#   baidu, boss/bosszhipin, doubao, taobao, alipay, jd, didi,
-#   douyin/tiktok, kuaishou, zhihu, ele, pinduoduo/pdd, gaode/amap
-#
-# Examples:
-#   bash phone_launch.sh wechat
-#   bash phone_launch.sh com.tencent.mm
-#   bash phone_launch.sh -l
-
-set -eo pipefail
-
-DEVICE=""
-QUIET=false
-LIST=false
-
-# Parse options manually to allow "phone_launch.sh <app> -d DEV" order
-ARGV=()
-i=1
-while [ $i -le $# ]; do
-    arg="${!i}"
-    case "$arg" in
-        -d) i=$((i+1)); DEVICE="${!i}" ;;
-        -l) LIST=true ;;
-        -q) QUIET=true ;;
-        -h|--help) head -21 "$0" | tail -16; exit 0 ;;
-        -*) echo "Unknown option: $arg" >&2; exit 1 ;;
-        *) ARGV+=("$arg") ;;
-    esac
-    i=$((i+1))
-done
-
-# App alias -> package name lookup (bash 3.2 compatible)
-resolve_app() {
-    local app
-    app=$(echo "$1" | tr '[:upper:]' '[:lower:]')
-    case "$app" in
-        wechat|weixin)       echo "com.tencent.mm" ;;
-        wecom|qiyeweixin)    echo "com.tencent.wework" ;;
-        meituan)             echo "com.sankuai.meituan" ;;
-        ctrip|xiecheng)      echo "ctrip.android.view" ;;
-        xiaohongshu|xhs)     echo "com.xingin.xhs" ;;
-        bilibili)            echo "tv.danmaku.bili" ;;
-        baidu)               echo "com.baidu.searchbox" ;;
-        boss|bosszhipin)     echo "com.hpbr.bosszhipin" ;;
-        doubao)              echo "com.larus.nova" ;;
-        taobao)              echo "com.taobao.taobao" ;;
-        alipay|zhifubao)     echo "com.eg.android.AlipayGphone" ;;
-        jd|jingdong)         echo "com.jingdong.app.mall" ;;
-        didi)                echo "com.sdu.didi.psnger" ;;
-        douyin|tiktok)       echo "com.ss.android.ugc.aweme" ;;
-        kuaishou)            echo "com.smile.gifmaker" ;;
-        zhihu)               echo "com.zhihu.android" ;;
-        ele|eleme)           echo "me.ele" ;;
-        pinduoduo|pdd)       echo "com.xunmeng.pinduoduo" ;;
-        gaode|amap)          echo "com.autonavi.minimap" ;;
-        qq)                  echo "com.tencent.mobileqq" ;;
-        wps)                 echo "cn.wps.moffice_eng" ;;
-        settings)            echo "com.android.settings" ;;
-        camera)              echo "com.android.camera" ;;
-        chrome)              echo "com.android.chrome" ;;
-        calendar)            echo "com.google.android.calendar" ;;
-        maps)                echo "com.google.android.apps.maps" ;;
-        phone)               echo "com.android.dialer" ;;
-        contacts)            echo "com.android.contacts" ;;
-        messages)            echo "com.google.android.apps.messaging" ;;
-        gallery)             echo "com.miui.gallery" ;;
-        files)               echo "com.android.fileexplorer" ;;
-        *)                   echo "" ;;
-    esac
-}
-
-list_aliases() {
-    cat <<'ALIASES'
-Known app aliases:
-  alipay/zhifubao  -> com.eg.android.AlipayGphone
-  amap/gaode       -> com.autonavi.minimap
-  baidu            -> com.baidu.searchbox
-  bilibili         -> tv.danmaku.bili
-  boss/bosszhipin  -> com.hpbr.bosszhipin
-  calendar         -> com.google.android.calendar
-  camera           -> com.android.camera
-  chrome           -> com.android.chrome
-  contacts         -> com.android.contacts
-  ctrip/xiecheng   -> ctrip.android.view
-  didi             -> com.sdu.didi.psnger
-  doubao           -> com.larus.nova
-  douyin/tiktok    -> com.ss.android.ugc.aweme
-  ele/eleme        -> me.ele
-  files            -> com.android.fileexplorer
-  gallery          -> com.miui.gallery
-  jd/jingdong      -> com.jingdong.app.mall
-  kuaishou         -> com.smile.gifmaker
-  maps             -> com.google.android.apps.maps
-  meituan          -> com.sankuai.meituan
-  messages         -> com.google.android.apps.messaging
-  phone            -> com.android.dialer
-  pinduoduo/pdd    -> com.xunmeng.pinduoduo
-  qq               -> com.tencent.mobileqq
-  settings         -> com.android.settings
-  taobao           -> com.taobao.taobao
-  wechat/weixin    -> com.tencent.mm
-  wecom/qiyeweixin -> com.tencent.wework
-  wps              -> cn.wps.moffice_eng
-  xiaohongshu/xhs  -> com.xingin.xhs
-  zhihu            -> com.zhihu.android
-ALIASES
-}
-
-if $LIST; then
-    list_aliases
-    exit 0
-fi
-
-if [ ${#ARGV[@]} -eq 0 ]; then
-    echo "Usage: $0 <app_name_or_package> [-d device] [-l] [-q]" >&2
-    exit 1
-fi
-
-APP_INPUT="${ARGV[0]}"
-
-ADB_CMD="adb"
-if [ -n "$DEVICE" ]; then
-    ADB_CMD="adb -s $DEVICE"
-fi
-
-info() { $QUIET || echo "[phone_launch] $*" >&2; }
-
-# Resolve package name
-PACKAGE=$(resolve_app "$APP_INPUT")
-if [ -n "$PACKAGE" ]; then
-    info "Resolved '$APP_INPUT' -> $PACKAGE"
-elif echo "$APP_INPUT" | grep -q '\.'; then
-    PACKAGE="$APP_INPUT"
-    info "Using package name directly: $PACKAGE"
-else
-    # Try fuzzy search on device
-    info "Unknown alias '$APP_INPUT', searching on device..."
-    MATCHES=$($ADB_CMD shell pm list packages | grep -i "$APP_INPUT" | sed 's/package://' || true)
-    if [ -z "$MATCHES" ]; then
-        echo "ERROR: No package found matching '$APP_INPUT'" >&2
-        echo "Use -l to list known aliases, or provide full package name" >&2
-        exit 1
-    fi
-    PACKAGE=$(echo "$MATCHES" | head -1 | tr -d '\r')
-    info "Found package: $PACKAGE"
-fi
-
-# Launch the app
-info "Launching $PACKAGE..."
-$ADB_CMD shell monkey -p "$PACKAGE" -c android.intent.category.LAUNCHER 1 >/dev/null 2>&1
-
-sleep 1
-info "Done."
-echo "$PACKAGE"

package/dist/builtin-skills/catalog/phone-adb-automation/phone_screenshot.sh

@@ -1,87 +0,0 @@
-#!/usr/bin/env bash
-# phone_screenshot.sh - Take a screenshot from connected Android phone via ADB
-#
-# Usage:
-#   bash phone_screenshot.sh [options]
-#
-# Options:
-#   -d DEVICE_ID    Target specific device (default: first connected device)
-#   -o OUTPUT_PATH  Output file path (default: /tmp/phone_screen.png)
-#   -s SCALE        Scale percentage 1-100 (default: 50)
-#   -r              Raw only - skip scaling, save native resolution
-#   -q              Quiet mode - suppress info messages
-#
-# Output:
-#   Saves scaled screenshot to OUTPUT_PATH
-#   Prints the output file path on success
-#
-# Examples:
-#   bash phone_screenshot.sh                      # Default: 50% scale to /tmp/phone_screen.png
-#   bash phone_screenshot.sh -o /tmp/my.png -s 40 # 40% scale to custom path
-#   bash phone_screenshot.sh -d abc123 -r         # Raw native resolution for specific device
-
-set -euo pipefail
-
-DEVICE=""
-OUTPUT="/tmp/phone_screen.png"
-SCALE=50
-RAW=false
-QUIET=false
-
-while getopts "d:o:s:rq" opt; do
-    case $opt in
-        d) DEVICE="$OPTARG" ;;
-        o) OUTPUT="$OPTARG" ;;
-        s) SCALE="$OPTARG" ;;
-        r) RAW=true ;;
-        q) QUIET=true ;;
-        *) echo "Usage: $0 [-d device] [-o output] [-s scale%] [-r] [-q]" >&2; exit 1 ;;
-    esac
-done
-
-ADB_CMD="adb"
-if [ -n "$DEVICE" ]; then
-    ADB_CMD="adb -s $DEVICE"
-fi
-
-info() { $QUIET || echo "[phone_screenshot] $*" >&2; }
-
-# Verify device connection
-if ! $ADB_CMD get-state >/dev/null 2>&1; then
-    echo "ERROR: No device connected or device unauthorized" >&2
-    echo "Run 'adb devices' to check connection status" >&2
-    exit 1
-fi
-
-REMOTE_PATH="/sdcard/screen_$(date +%s).png"
-RAW_PATH="/tmp/phone_raw_$$.png"
-
-info "Capturing screenshot..."
-$ADB_CMD shell screencap -p "$REMOTE_PATH"
-
-info "Pulling to local..."
-$ADB_CMD pull "$REMOTE_PATH" "$RAW_PATH" >/dev/null 2>&1
-
-info "Cleaning up device temp file..."
-$ADB_CMD shell rm -f "$REMOTE_PATH"
-
-if $RAW; then
-    mv "$RAW_PATH" "$OUTPUT"
-    info "Saved raw screenshot to $OUTPUT"
-else
-    # Get native resolution from the image
-    NATIVE_W=$(sips -g pixelWidth "$RAW_PATH" | tail -1 | awk '{print $2}')
-    NATIVE_H=$(sips -g pixelHeight "$RAW_PATH" | tail -1 | awk '{print $2}')
-
-    SCALED_W=$((NATIVE_W * SCALE / 100))
-    SCALED_H=$((NATIVE_H * SCALE / 100))
-
-    info "Native: ${NATIVE_W}x${NATIVE_H} -> Scaled (${SCALE}%): ${SCALED_W}x${SCALED_H}"
-    sips --resampleHeightWidth "$SCALED_H" "$SCALED_W" "$RAW_PATH" --out "$OUTPUT" >/dev/null 2>&1
-
-    rm -f "$RAW_PATH"
-    info "Saved scaled screenshot to $OUTPUT"
-fi
-
-# Output the path for easy capture by calling scripts
-echo "$OUTPUT"

package/dist/builtin-skills/catalog/phone-adb-automation/phone_security_kbd.py

@@ -1,174 +0,0 @@
-#!/usr/bin/env python3
-"""
-phone_security_kbd.py - Handle randomized security keyboards on Android.
-
-Financial apps (banking, stock trading) use security keyboards where digit
-positions change every time. This script reads the UI dump to map digit
-positions and types a sequence by tapping each digit.
-
-Usage:
-    python3 phone_security_kbd.py <digits> [options]
-
-Options:
-    -d DEVICE_ID    Target specific device
-    -f XML_FILE     Use existing UI dump XML file
-    --delay MS      Delay between taps in ms (default: 200)
-    --dry-run       Show mapped positions without tapping
-    -q              Quiet mode
-
-How it works:
-    1. Dumps the UI hierarchy via uiautomator
-    2. Finds all clickable elements with numeric content-desc (0-9)
-    3. Maps each digit to its center coordinates
-    4. Taps each digit in sequence with configurable delay
-
-Examples:
-    python3 phone_security_kbd.py "123456"              # Type PIN 123456
-    python3 phone_security_kbd.py "888888" --dry-run    # Show positions only
-    python3 phone_security_kbd.py "1234" --delay 300    # Slower tapping
-    python3 phone_security_kbd.py "0000" -f /tmp/ui.xml # Use existing dump
-"""
-
-import argparse
-import os
-import re
-import subprocess
-import sys
-import tempfile
-import time
-import xml.etree.ElementTree as ET
-
-
-def run_adb(device, *args):
-    cmd = ["adb"]
-    if device:
-        cmd += ["-s", device]
-    cmd += list(args)
-    result = subprocess.run(cmd, capture_output=True, text=True)
-    return result.stdout, result.stderr, result.returncode
-
-
-def dump_ui(device=None):
-    remote_path = "/sdcard/ui_kbd.xml"
-    local_path = os.path.join(tempfile.gettempdir(), f"phone_kbd_{os.getpid()}.xml")
-
-    _, stderr, rc = run_adb(device, "shell", "uiautomator", "dump", remote_path)
-    if rc != 0:
-        print(f"ERROR: uiautomator dump failed: {stderr}", file=sys.stderr)
-        sys.exit(1)
-
-    _, stderr, rc = run_adb(device, "pull", remote_path, local_path)
-    if rc != 0:
-        print(f"ERROR: Failed to pull UI dump: {stderr}", file=sys.stderr)
-        sys.exit(1)
-
-    run_adb(device, "shell", "rm", "-f", remote_path)
-    return local_path
-
-
-def parse_keyboard_map(xml_path):
-    """Parse UI dump and return dict of digit/char -> (center_x, center_y)."""
-    tree = ET.parse(xml_path)
-    root = tree.getroot()
-    kbd_map = {}
-
-    for node in root.iter():
-        desc = node.get("content-desc", "").strip()
-        text = node.get("text", "").strip()
-        bounds = node.get("bounds", "")
-        clickable = node.get("clickable", "false")
-
-        if not bounds or clickable != "true":
-            continue
-
-        # Use content-desc first, fallback to text
-        label = desc or text
-        if not label:
-            continue
-
-        m = re.findall(r"\d+", bounds)
-        if len(m) != 4:
-            continue
-
-        x1, y1, x2, y2 = int(m[0]), int(m[1]), int(m[2]), int(m[3])
-        cx = (x1 + x2) // 2
-        cy = (y1 + y2) // 2
-
-        # Map single digits and common labels
-        if label in "0123456789":
-            kbd_map[label] = (cx, cy)
-        elif label.lower() in ("delete", "backspace", "del", "删除"):
-            kbd_map["delete"] = (cx, cy)
-        elif label.lower() in ("confirm", "ok", "done", "确定", "确认", "完成"):
-            kbd_map["confirm"] = (cx, cy)
-        elif label == ".":
-            kbd_map["."] = (cx, cy)
-
-    return kbd_map
-
-
-def tap_digit(device, x, y):
-    run_adb(device, "shell", "input", "tap", str(x), str(y))
-
-
-def main():
-    parser = argparse.ArgumentParser(description="Type on security keyboard")
-    parser.add_argument("digits", help="Digits to type (e.g., '123456')")
-    parser.add_argument("-d", "--device", help="Target device ID")
-    parser.add_argument("-f", "--file", help="Existing UI dump XML")
-    parser.add_argument("--delay", type=int, default=200, help="Delay between taps (ms)")
-    parser.add_argument("--dry-run", action="store_true", help="Show positions only")
-    parser.add_argument("-q", "--quiet", action="store_true", help="Quiet mode")
-    args = parser.parse_args()
-
-    info = lambda msg: None if args.quiet else print(f"[security_kbd] {msg}", file=sys.stderr)
-
-    # Get UI dump
-    if args.file:
-        xml_path = args.file
-        cleanup = False
-    else:
-        info("Dumping UI hierarchy...")
-        xml_path = dump_ui(args.device)
-        cleanup = True
-
-    # Parse keyboard layout
-    kbd_map = parse_keyboard_map(xml_path)
-
-    if cleanup:
-        os.unlink(xml_path)
-
-    if not kbd_map:
-        print("ERROR: No keyboard digits found in UI dump", file=sys.stderr)
-        print("Make sure a security keyboard is visible on screen", file=sys.stderr)
-        sys.exit(1)
-
-    info(f"Keyboard map: { {k: v for k, v in sorted(kbd_map.items()) if k in '0123456789'} }")
-
-    # Validate all digits are mappable
-    for ch in args.digits:
-        if ch not in kbd_map:
-            print(f"ERROR: Digit '{ch}' not found on keyboard. Available: {sorted(kbd_map.keys())}", file=sys.stderr)
-            sys.exit(1)
-
-    if args.dry_run:
-        print("Dry run - would tap:")
-        for ch in args.digits:
-            x, y = kbd_map[ch]
-            print(f"  '{ch}' -> ({x}, {y})")
-        return
-
-    # Type each digit
-    for i, ch in enumerate(args.digits):
-        x, y = kbd_map[ch]
-        info(f"Tapping '{ch}' at ({x}, {y})")
-        tap_digit(args.device, x, y)
-        if i < len(args.digits) - 1:
-            time.sleep(args.delay / 1000.0)
-
-    info(f"Typed {len(args.digits)} digits successfully")
-    print("kbd_input_ok")
-
-
-if __name__ == "__main__":
-    main()