franzenzenhofer/big-arrow-on-the-screen

What it solves

AI agents often struggle to communicate with humans when they need a specific user action on a screen (e.g., "click Allow in the dialog"). Instead of printing a text message in a terminal that the user might not be looking at, big-arrow-on-the-screen allows agents to visually point to specific UI elements, windows, or coordinates on a macOS desktop with highly visible arrows and labels.

How it works

It is a macOS command-line tool written in Swift that creates a transparent overlay window on top of all other windows, displays, and Spaces. It draws arrows, boxes, or rings that point to a target—which can be defined by a label, coordinate, or window title—without stealing keyboard focus or blocking clicks to the target element. It includes a "skill" configuration for AI agents (like Claude Code and Codex) to enable them to use these commands autonomously.

Who it’s for

• AI Agent Developers: To give their agents a way to guide users through UI interactions.
• End Users: Who are being guided by an agent through complex settings or permissions.
• Developers/Testers: For creating visual demos, documentation, or debugging UI coordinates.

Highlights

• Non-Intrusive: Clicks pass through to the target; keyboard focus remains on the current app.
• Flexible Targeting: Point by UI element label, global coordinates, window title, or specific browser tabs.
• Rich Visuals: Supports various shapes (straight, zigzag, spiral), styles (arrow, ring, box), sizes, and custom colors.
• Interactive Elements: Includes a copy-to-clipboard button within the sign and an optional close button.
• Agent-Ready: Provides a dedicated skill for LLM agents and uses standard exit codes for programmatic feedback.
• Zero-Permission Drawing: Drawing the overlay requires no macOS permissions, though finding specific elements may require Accessibility permissions.