Function createCliExecutor

src/utils/computerUse/executor.ts:259–645

(opts: {
  getMouseAnimationEnabled: () => boolean
  getHideBeforeActionEnabled: () => boolean
})

// ── Factory ───────────────────────────────────────────────────────────────

export function createCliExecutor(opts: {
  getMouseAnimationEnabled: () => boolean
  getHideBeforeActionEnabled: () => boolean
}): ComputerExecutor {
  if (process.platform !== 'darwin') {
    throw new Error(
      `createCliExecutor called on ${process.platform}. Computer control is macOS-only.`,
    )
  }

  // Swift loaded once at factory time — every executor method needs it.
  // Input loaded lazily via requireComputerUseInput() on first mouse/keyboard
  // call — it caches internally, so screenshot-only flows never pull the
  // enigo .node.
  const cu = requireComputerUseSwift()

  const { getMouseAnimationEnabled, getHideBeforeActionEnabled } = opts
  const terminalBundleId = getTerminalBundleId()
  const surrogateHost = terminalBundleId ?? CLI_HOST_BUNDLE_ID
  // Swift 0.2.1's captureExcluding/captureRegion take an ALLOW list despite the
  // name (apps#30355 — complement computed Swift-side against running apps).
  // The terminal isn't in the user's grants so it's naturally excluded, but if
  // the package ever passes it through we strip it here so the terminal never
  // photobombs a screenshot.
  const withoutTerminal = (allowed: readonly string[]): string[] =>
    terminalBundleId === null
      ? [...allowed]
      : allowed.filter(id => id !== terminalBundleId)

  logForDebugging(
    terminalBundleId
      ? `[computer-use] terminal ${terminalBundleId} → surrogate host (hide-exempt, activate-skip, screenshot-excluded)`
      : '[computer-use] terminal not detected; falling back to sentinel host',
  )

  return {
    capabilities: {
      ...CLI_CU_CAPABILITIES,
      hostBundleId: CLI_HOST_BUNDLE_ID,
    },

    // ── Pre-action sequence (hide + defocus) ────────────────────────────

    async prepareForAction(
      allowlistBundleIds: string[],
      displayId?: number,
    ): Promise<string[]> {
      if (!getHideBeforeActionEnabled()) {
        return []
      }
      // prepareDisplay isn't @MainActor (plain Task{}), but its .hide() calls
      // trigger window-manager events that queue on CFRunLoop. Without the
      // pump, those pile up during Swift's ~1s of usleeps and flush all at
      // once when the next pumped call runs — visible window flashing.
      // Electron drains CFRunLoop continuously so Cowork doesn't see this.
      // Worst-case 100ms + 5×200ms safety-net ≈ 1.1s, well under the 30s
      // drainRunLoop ceiling.
      //