Skill Recorder is a desktop application built with TypeScript and Electron that captures on-screen work sessions and transforms them into reusable AI agent skills and automations. The core workflow involves recording a single task execution, analyzing it with GitHub Copilot CLI to extract intent and ordered steps, and then generating either a Skill (a procedure an agent runs on demand) or an Automation (the same procedure triggered on a schedule). The application targets both macOS as its primary platform and Windows 11 in x64 and ARM64 variants.
The recording process captures screen activity, window and application switches, browser URLs on macOS, clipboard previews, and optional spoken narration transcribed on-device using Whisper's 99 supported languages. All capture, storage, frame extraction, and transcription occur locally on the user's computer. Only when the user explicitly chooses to Analyze does the application send the event timeline, extracted screen images, and narration text to GitHub's cloud for Copilot processing. The application includes prominent warnings against recording passwords, tokens, API keys, or other confidential information.
The user interface consists of a record button with keyboard shortcuts (Command-Shift-R on macOS, Ctrl-Shift-R on Windows), a timer, optional narration toggle with language and microphone settings, and readiness checks. During recording, an always-on-top control bar displays capture and microphone state, allowing users to mute, unmute, or switch microphones on the fly. After recording, users can review and edit the reconstructed intent and steps before approving the analysis for skill or automation generation.
Installation follows a source-release model where a single command downloads a pinned Node.js runtime, builds the exact release commit on the user's machine, and adds a Skill Recorder application without global installation. The application requires a GitHub account with Copilot access, and the Copilot CLI ships with the app. On macOS, the installation adds a Skill Recorder app to the Applications folder accessible via Spotlight, Launchpad, or the Dock. On Windows, it creates desktop and Start Menu shortcuts.
The generated skills and automations prefer the agent's native tools such as the gh CLI or web_fetch over replaying UI clicks, and they generalize from single recorded examples. This means recording one form submission can teach an agent to submit multiple forms. The application integrates with Microsoft Scout, Microsoft Copilot Cowork, and Copilot Studio as target platforms for the created skills and automations.
Development requires Node.js 24, and the repository includes npm scripts for local development with Vite and hot-reload Electron launching. The project maintains fixture-based evaluation suites for the Copilot describer and builders, documented in an evals directory. The codebase includes comprehensive documentation covering installation options, release procedures, evaluation harness details, and third-party licenses. Security vulnerabilities should be reported through Microsoft's coordinated-disclosure process rather than public GitHub issues. The project is licensed under MIT and welcomes contributions subject to a Contributor License Agreement.
