eli0shin/.dotfiles/agent-browser

agent-browser

Fast browser automation CLI for AI agents. Chrome/Chromium via CDP with accessibility-tree snapshots and compact @eN element refs.

Install: npm i -g agent-browser && agent-browser install

Local browser rules

These rules override the CLI-provided skills:

• For authenticated tasks, launch the user's installed Google Chrome with their profile: use agent-browser profiles to find the Chrome profile, then start a named, isolated session with --profile <name> --executable-path <installed Google Chrome binary>. The binary's path is OS-specific; locate it on the current machine. agent-browser's bundled Chromium opens the copied profile signed out, so real Chrome is what carries the user's logins. This copies the profile to a temporary directory and does not control the visible Chrome window. Ask the user to select the profile when the choice is ambiguous. A signed-out page means the launch is wrong: fix the browser or profile and relaunch, since the profile supplies the login.
• Derive one stable session ID with agent-browser session id --scope worktree --prefix <task> and pass --session <id> on every browser command.
• Work only in the session's active tab. Use get url and snapshot -i to inspect it.
• Never list tabs. Do not run `tab list`, including for diagnostics or recovery. Open a new isolated session if the active tab cannot be recovered.
• Never use `--auto-connect` or `--cdp`. They can attach to the user's visible, authenticated browser and disrupt it.

Start here

Before running any other agent-browser command, load the version-matched workflow content:

agent-browser skills get core
agent-browser skills get core --full      # only when the full command reference is necessary

Apply the local browser rules above if the CLI-provided content conflicts with them.

Specialized skills

Load a specialized skill when the task falls outside browser web pages:

agent-browser skills get electron          # Electron desktop apps (VS Code, Slack, Discord, Figma, ...)
agent-browser skills get slack             # Slack workspace automation
agent-browser skills get dogfood           # Exploratory testing / QA / bug hunts
agent-browser skills get vercel-sandbox    # agent-browser inside Vercel Sandbox microVMs
agent-browser skills get agentcore         # AWS Bedrock AgentCore cloud browsers

Run agent-browser skills list to see everything available on the installed version.