Automate macOS UI interactions with an observe-act-verify loop and JSON output.
npx skills add https://github.com/wenerme/ai --skill peekaboo-wenerme
Please help me install this Agent Skill. Skill: peekaboo Source: https://github.com/wenerme/ai/tree/main/skills/peekaboo Command: npx skills add https://github.com/wenerme/ai --skill peekaboo-wenerme
Peekaboo automates macOS user interface tasks like clicking, typing, window/menu navigation, permissions setup, and UI inspection so you don’t have to perform repetitive manual steps.
list and see to locate apps, windows, screens, menus, and interactive elements (with refreshed element IDs after changes) for reliable targeting.Ask the AI agent to use Peekaboo to open the target app, observe the relevant UI with a fresh element map, perform the required click or typing action using element IDs, and then verify the result with a new see or list output.
High-intent search queries and answers about installing and using this skill.
macOS UI automation can be performed by inspecting UI state to locate interactive elements, executing targeted actions using element IDs, and verifying outcomes through an observe-act-verify loop. This non-coordinate-first approach reduces flaky automation by dynamically refreshing element IDs.
To troubleshoot macOS automation failures, check the TCC-aware permissions status and Bridge host status. Diagnosing these requirements helps resolve failures related to accessibility permissions and host or remote execution differences.
Yes, you can run macOS automation as an MCP server to handle app, window, menu, and dialog control for agents. It provides agent-oriented commands for higher-level automation while respecting safety rules for destructive actions.
You locate UI elements by using discovery commands to list apps, windows, screens, and menus. After identifying the target, you interact by clicking, typing, or pressing hotkeys, then verify the result with a new inspection output.
macOS UI automation fails during remote execution due to Bridge host differences and unmet TCC permissions. Checking the Bridge status and verifying accessibility permissions will diagnose these host-specific execution failures.
rp4riDiagnose and recover stalled Claude Code worker sessions in multi-agent orchestration.
openaeonAutomate macOS desktop UI capture and interaction via the Peekaboo CLI.
vuminhieuDiscover, analyze, and execute MCP capabilities across configured servers.
the-open-agentAutomate macOS UI interactions via CLI commands for testing and scripting.