The Agent Skills Directory

[COMMAND_EXECUTION]: The terminal automation feature utilizes the node-pty library to spawn a real, interactive shell (defaulting to /bin/zsh or the user's $SHELL) on the host machine, which allows for the execution of arbitrary system commands as part of demo scenarios.
[DATA_EXFILTRATION]: The init flow instructions direct the agent to perform extensive codebase exploration, specifically targeting .env.example and .env.local files to discover sensitive information such as API endpoints, URLs, and authentication secrets.
[REMOTE_CODE_EXECUTION]: The skill's automation engine includes an exec step that allows for the execution of arbitrary asynchronous JavaScript code within the automation context.
[REMOTE_CODE_EXECUTION]: The render-page.js module implements a frame-by-frame rendering pipeline that uses eval() within a headless Playwright browser to execute dynamically generated Canvas2D drawing commands for video annotations.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of multiple external dependencies and binaries using npm and npx, including the Playwright browser engine, the node-pty library, and a vendor-specific library via npx skills add llmer/skill-demoflow.
[COMMAND_EXECUTION]: The studio.js component launches a local web server and WebSocket server on port 3274, which listens for network connections and serves files from the skill's output directory to the local network.

llmer-demo