What problem does it solve? Automating applications that expose no API requires driving their user interface directly, and naive UI automation breaks silently through sleeps, pixel coordinates, unverified writes, and unbounded loops. This Skill guides the authoring of Keyhac actions that read and drive another application's accessibility tree reliably on macOS and Windows. ## Core Features & Use Cases - Accessibility-tree automation: Generate plain-Python Keyhac actions that locate elements by identifier, name, role, or capability, act through verified verbs with explicit preconditions and postconditions, and never rely on sleeps or pixel coordinates. - Structured extraction and form filling: Read tables and paginated results, fill forms from CSV data with verified writes, handle dialogs with appear-act-gone sequencing, and checkpoint progress so interrupted runs resume safely. - Platform quirk guidance: Reference documentation covers measured macOS and Windows behaviors such as Chromium content access, role vocabulary differences, and toggle-state handling. - Use Case: Ask the assistant to extract every row from a paginated results table in Safari into a CSV; it inspects the live screen, writes a ThreadedAction module into ~/.keyhac/extensions/, runs it via the Keyhac MCP tools, reads the result, and hands back a key-binding snippet. ## Quick Start Ask the assistant to write a Keyhac action that extracts the records from the currently open application window and saves them to a CSV on your Desktop.
