@syntesseraai/opencode-feature-factory 0.2.44 → 0.2.45

package/package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "@syntesseraai/opencode-feature-factory",
-  "version": "0.2.44",
+  "version": "0.2.45",
   "type": "module",
   "description": "OpenCode plugin for Feature Factory agents - provides sub-agents and skills for validation, review, security, and architecture assessment",
   "license": "MIT",
@@ -30,6 +30,7 @@
     "security-audit",
     "aws-well-architected"
   ],
+  "scripts": {},
   "dependencies": {
     "@opencode-ai/plugin": "^1.1.48",
     "glob": "^10.0.0",
@@ -39,6 +40,5 @@
     "@types/bun": "^1.2.6",
     "@types/node": "^22.0.0",
     "typescript": "^5.0.0"
-  },
-  "scripts": {}
-}
+  }
+}

package/skills/ff-computer-use/SKILL.md

@@ -1,473 +0,0 @@
-# ff-computer-use: iPhone Mirroring Interaction Skill
-
-> **Purpose:** Enable agents to interact with iPhone apps via macOS iPhone Mirroring using MCP tools for mouse control, keyboard input, screenshots, and scrolling.
-
-## Required MCP Servers
-
-This skill depends on two MCP servers that must be configured:
-
-### 1. `computer-use-mcp` (Primary — mouse, keyboard, screenshots)
-
-Provides the `computer` tool for clicking, dragging, typing, screenshots, and cursor control.
-
-```json
-"computer-use-mcp": {
-  "type": "local",
-  "command": ["npx", "-y", "computer-use-mcp@latest"],
-  "enabled": true
-}
-```
-
-### 2. `macos-automator` (Supplementary — scrolling, AppleScript, window discovery)
-
-Provides `execute_script` for running AppleScript/JXA (needed for scroll wheel events and window management).
-
-```json
-"macos-automator": {
-  "type": "local",
-  "command": ["npx", "-y", "@steipete/macos-automator-mcp@latest"],
-  "enabled": true
-}
-```
-
-## Prerequisites
-
-- **macOS 15+ (Sequoia)** with iPhone Mirroring enabled
-- **iPhone Mirroring app** running and connected to your iPhone
-- **Accessibility permissions** granted to the terminal/editor app (System Settings → Privacy & Security → Accessibility)
-- **Screen Recording permissions** granted to the terminal/editor app (System Settings → Privacy & Security → Screen Recording)
-- **Node.js** installed (for `npx` to run the MCP servers)
-
----
-
-## Architecture
-
-### How It Works
-
-iPhone Mirroring on macOS renders the iPhone screen as a native macOS window. We interact with it using:
-
-1. **`computer` tool** (computer-use-mcp) — Mouse clicks, drags, typing, key presses, and screenshots via nut.js
-2. **`execute_script` tool** (macos-automator) — AppleScript/JXA for scroll wheel events (CGEvent), window discovery, and app management
-3. **`get_scripting_tips` tool** (macos-automator) — Search 200+ pre-built macOS automation scripts
-
-### Why Two MCPs?
-
-- **`computer-use-mcp`** handles most interactions (click, drag, type, screenshot) but has **no scroll wheel action**
-- **`macos-automator`** fills the gap with JXA scripts that post `CGEvent` scroll wheel events, which iPhone Mirroring correctly translates to touch scrolls
-
-### What Doesn't Work
-
-- **Arrow keys for scrolling** — iPhone Mirroring doesn't translate keyboard arrow keys to scrolling
-- **Mouse drag for scrolling** — Simple mouse drag doesn't trigger touch-scroll in iPhone Mirroring
-- **mobile-mcp / Appium** — External mobile automation tools don't work through the mirroring layer
-- **AX UI inspection** — iPhone Mirroring window doesn't expose iOS accessibility elements to macOS
-
----
-
-## Tool Reference
-
-### `computer` — Mouse, Keyboard & Screenshots
-
-The `computer` tool from `computer-use-mcp` provides these actions:
-
-#### `get_screenshot` — Capture the Screen
-
-Takes a screenshot and returns it as a base64 PNG with display dimensions.
-
-```
-Tool: computer
-Action: get_screenshot
-```
-
-> **Tip:** Use this before and after every interaction to understand the current state.
-
-#### `left_click` — Click at Coordinates
-
-```
-Tool: computer
-Action: left_click
-Coordinate: [x, y]
-```
-
-#### `right_click` — Right-Click at Coordinates
-
-```
-Tool: computer
-Action: right_click
-Coordinate: [x, y]
-```
-
-#### `double_click` — Double-Click at Coordinates
-
-```
-Tool: computer
-Action: double_click
-Coordinate: [x, y]
-```
-
-#### `mouse_move` — Move Cursor
-
-```
-Tool: computer
-Action: mouse_move
-Coordinate: [x, y]
-```
-
-#### `drag` — Drag to Coordinates
-
-Presses left mouse button at current position, moves to target, and releases.
-
-```
-Tool: computer
-Action: drag
-Coordinate: [x, y]    # Target/end position
-```
-
-> **Important:** Move the cursor to the start position first with `mouse_move`, then use `drag` to the end position.
-
-#### `type` — Type Text
-
-Types a string of text. Use this when a text field is focused.
-
-```
-Tool: computer
-Action: type
-Text: "Hello, World!"
-```
-
-#### `key` — Press Keyboard Shortcut
-
-Press a key or key combination.
-
-```
-Tool: computer
-Action: key
-Text: "Return"          # Enter key
-Text: "BackSpace"       # Delete
-Text: "Escape"          # Escape
-Text: "Tab"             # Tab
-Text: "ctrl+a"          # Select all
-Text: "cmd+c"           # Copy
-```
-
-#### `cursor_position` — Get Current Cursor Position
-
-```
-Tool: computer
-Action: cursor_position
-```
-
----
-
-### `execute_script` — Scrolling & Window Management
-
-The `execute_script` tool from `macos-automator` runs AppleScript or JXA scripts.
-
-#### Scroll via CGEvent (JXA)
-
-This is the **only reliable way** to scroll in iPhone Mirroring:
-
-```javascript
-// Scroll DOWN (content moves up, see content below)
-// Use execute_script with language: "javascript"
-ObjC.import('CoreGraphics');
-
-const x = CURSOR_X; // Absolute X coordinate
-const y = CURSOR_Y; // Absolute Y coordinate
-const delta = -5; // Negative = scroll down, Positive = scroll up
-
-// Move cursor to position
-const moveEvent = $.CGEventCreateMouseEvent(null, $.kCGEventMouseMoved, $.CGPointMake(x, y), 0);
-$.CGEventPost($.kCGHIDEventTap, moveEvent);
-delay(0.05);
-
-// Post scroll wheel event
-const scrollEvent = $.CGEventCreateScrollWheelEvent(null, 1, 1, delta);
-$.CGEventPost($.kCGHIDEventTap, scrollEvent);
-```
-
-**Scroll direction:**
-
-- `delta > 0` → Scroll **UP** (content moves down, you see content above)
-- `delta < 0` → Scroll **DOWN** (content moves up, you see content below)
-
-**Recommended scroll amounts:**
-
-- Small scroll: `3` to `5`
-- Medium scroll: `8` to `12`
-- Large scroll: `15` to `25`
-- Full page: `30` to `50`
-
-#### Get Window Bounds (AppleScript)
-
-```applescript
--- Use execute_script with language: "applescript"
-tell application "System Events"
-    tell process "iPhone Mirroring"
-        set winPos to position of window 1
-        set winSize to size of window 1
-        return (item 1 of winPos as text) & "," & (item 2 of winPos as text) & "," & (item 1 of winSize as text) & "," & (item 2 of winSize as text)
-    end tell
-end tell
-```
-
-Returns: `x,y,width,height` — where `(x, y)` is the top-left corner of the window.
-
-#### Activate iPhone Mirroring (AppleScript)
-
-```applescript
-tell application "iPhone Mirroring" to activate
-delay 2
-```
-
-#### Bring to Front (AppleScript)
-
-```applescript
-tell application "System Events"
-    tell process "iPhone Mirroring"
-        set frontmost to true
-    end tell
-end tell
-```
-
----
-
-### `get_scripting_tips` — Find Pre-Built Scripts
-
-Search the macos-automator knowledge base for automation scripts:
-
-```
-Tool: get_scripting_tips
-search_term: "screenshot"
-limit: 5
-```
-
----
-
-## Coordinate System
-
-### Understanding Coordinates
-
-All coordinates are **absolute screen coordinates** (macOS global coordinate space).
-
-The iPhone Mirroring window has:
-
-- A **title bar** (~28px) at the top that is NOT part of the iPhone screen
-- The **iPhone content area** below the title bar
-
-### Coordinate Calculation
-
-To interact with a point within the iPhone screen:
-
-```
-absoluteX = windowX + relativeX
-absoluteY = windowY + titleBarHeight + relativeY
-```
-
-Where:
-
-- `windowX`, `windowY` = window position from AppleScript (see Get Window Bounds above)
-- `titleBarHeight` = ~28 pixels (macOS window title bar)
-- `relativeX`, `relativeY` = position within the iPhone content area
-
-### Typical iPhone Mirroring Window Dimensions
-
-| iPhone Model    | Window Width | Content Height | Total Height (with title bar) |
-| --------------- | ------------ | -------------- | ----------------------------- |
-| Standard (6.1") | ~336px       | ~728px         | ~756px                        |
-| Plus/Max (6.7") | ~336px       | ~728px         | ~756px                        |
-
-> **Note:** Actual dimensions may vary. Always discover dynamically using the Get Window Bounds AppleScript.
-
-### Common Tap Targets
-
-For a standard iPhone Mirroring window at position (x, y):
-
-| Target                    | Approximate Coordinates   |
-| ------------------------- | ------------------------- |
-| Status bar                | `(x + 168, y + 28 + 10)`  |
-| Center of screen          | `(x + 168, y + 28 + 364)` |
-| Bottom tab bar (1st item) | `(x + 42, y + 28 + 695)`  |
-| Bottom tab bar (2nd item) | `(x + 126, y + 28 + 695)` |
-| Bottom tab bar (3rd item) | `(x + 210, y + 28 + 695)` |
-| Bottom tab bar (4th item) | `(x + 294, y + 28 + 695)` |
-| Home indicator area       | `(x + 168, y + 28 + 720)` |
-| Back button (top-left)    | `(x + 30, y + 28 + 55)`   |
-| Navigation title          | `(x + 168, y + 28 + 55)`  |
-
----
-
-## Workflows
-
-### Workflow 1: Discover and Interact
-
-The standard workflow for interacting with an iPhone app:
-
-```
-Step 1: Activate iPhone Mirroring
-→ execute_script (applescript): tell application "iPhone Mirroring" to activate
-
-Step 2: Get window position
-→ execute_script (applescript): Get Window Bounds script (see above)
-→ Parse the returned "x,y,width,height" string
-
-Step 3: Take a screenshot to see current state
-→ computer: get_screenshot
-
-Step 4: Analyze the screenshot
-→ Identify UI elements and calculate their absolute coordinates
-→ Remember: absoluteY = windowY + 28 + relativeY
-
-Step 5: Click on a target
-→ computer: left_click at [absoluteX, absoluteY]
-
-Step 6: Wait briefly, then screenshot to verify
-→ (pause ~1 second)
-→ computer: get_screenshot
-```
-
-### Workflow 2: Scroll Through Content
-
-```
-Step 1: Get window position (see Workflow 1, Steps 1-2)
-→ Calculate center: centerX = windowX + width/2, centerY = windowY + 28 + contentHeight/2
-
-Step 2: Scroll down
-→ execute_script (javascript): CGEvent scroll script with x=centerX, y=centerY, delta=-5
-
-Step 3: Wait and screenshot
-→ (pause ~0.5 seconds)
-→ computer: get_screenshot
-
-Step 4: Repeat as needed with larger/smaller delta values
-```
-
-### Workflow 3: Navigate Between Screens
-
-```
-Step 1: Click a list item to navigate forward
-→ computer: left_click at [itemX, itemY]
-
-Step 2: Wait for transition
-→ (pause ~1 second)
-
-Step 3: Screenshot the new screen
-→ computer: get_screenshot
-
-Step 4: Go back (tap back button, typically top-left)
-→ computer: left_click at [windowX + 30, windowY + 28 + 55]
-```
-
-### Workflow 4: Type Text
-
-iPhone Mirroring supports keyboard input when a text field is focused:
-
-```
-Step 1: Tap on a text field to focus it
-→ computer: left_click at [fieldX, fieldY]
-
-Step 2: Wait for keyboard to appear
-→ (pause ~0.5 seconds)
-
-Step 3: Type the text
-→ computer: type "Hello, World!"
-
-Step 4: Press Enter if needed
-→ computer: key "Return"
-```
-
-### Workflow 5: Swipe / Drag Gesture
-
-```
-Step 1: Move cursor to start position
-→ computer: mouse_move to [startX, startY]
-
-Step 2: Drag to end position
-→ computer: drag to [endX, endY]
-
-Examples:
-- Swipe left (next page): mouse_move [300, 500] → drag [100, 500]
-- Pull to refresh: mouse_move [168, 400] → drag [168, 600]
-- Swipe right (go back): mouse_move [50, 400] → drag [300, 400]
-```
-
----
-
-## Troubleshooting
-
-### "Not permitted to send input events"
-
-Your terminal/editor needs Accessibility permissions:
-
-1. Open **System Settings → Privacy & Security → Accessibility**
-2. Add your terminal app (Terminal.app, iTerm2, VS Code, Cursor, etc.)
-3. Toggle it ON
-4. Restart the app
-
-### Screenshots not capturing correctly
-
-Your terminal/editor needs Screen Recording permissions:
-
-1. Open **System Settings → Privacy & Security → Screen Recording**
-2. Add your terminal/editor app
-3. Toggle it ON
-4. Restart the app
-
-### iPhone Mirroring window not found
-
-```
-→ execute_script (applescript):
-tell application "System Events" to name of every process whose name contains "iPhone"
-```
-
-If not running:
-
-```
-→ execute_script (applescript):
-tell application "iPhone Mirroring" to activate
-delay 3
-```
-
-### Clicks not registering
-
-1. Ensure the iPhone Mirroring window is **not minimized**
-2. Ensure coordinates are within the window bounds
-3. Bring the window to focus first:
-   ```
-   → execute_script (applescript):
-   tell application "iPhone Mirroring" to activate
-   ```
-
-### Scroll not working
-
-1. Ensure the cursor position is within the iPhone content area (not the title bar)
-2. Try larger scroll values (e.g., `-15` instead of `-3`)
-3. Add a small delay between scroll events if doing multiple scrolls
-4. Verify you're using the JXA CGEvent scroll approach (not arrow keys or mouse drag)
-
----
-
-## Best Practices
-
-1. **Always screenshot first** — Before interacting, take a screenshot to understand the current state
-2. **Analyze screenshots visually** — Use the returned screenshot image to identify UI elements and their positions
-3. **Calculate coordinates dynamically** — Always get window position via AppleScript; never hardcode coordinates
-4. **Add delays after interactions** — Wait ~1 second after clicks/scrolls to let the UI update before screenshotting
-5. **Scroll incrementally** — Use small scroll values (5-10) and check results rather than large jumps
-6. **Verify after each action** — Take a screenshot after each interaction to confirm it worked
-7. **Handle the title bar** — Always add 28px to Y coordinates to account for the macOS title bar
-8. **Keep iPhone Mirroring focused** — Activate the app before interactions
-9. **Use `mouse_move` before `drag`** — The drag action goes FROM current cursor position TO the target
-10. **Prefer `computer` tool for most actions** — Only use `execute_script` for scrolling and window management
-
----
-
-## Limitations
-
-- **No element inspection** — Cannot query iOS accessibility tree through mirroring
-- **Coordinate-based only** — All interactions require knowing pixel coordinates
-- **Single touch only** — Cannot simulate multi-touch gestures (pinch, rotate)
-- **No gesture recognition** — Drag works but complex gestures may not translate correctly
-- **Screen resolution dependent** — Coordinates depend on display scaling settings
-- **Requires visual analysis** — Must use screenshots + vision to understand UI state
-- **No scroll in `computer` tool** — Must use `execute_script` with JXA CGEvent for scrolling