browse

What This Skill Does

The browse skill provides an end-to-end framework for developing and deploying robust browser automation agents using the Stagehand CLI and Browserbase platform. It allows users to interact with websites programmatically, scrape dynamic data, and execute complex workflows like form submission, navigation, and state verification. By leveraging the Stagehand ecosystem, this skill simplifies the transition from local exploratory browsing to reliable, cloud-deployed automation functions.

Installation

To integrate this capability into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/pkiv/browse. Once installed, ensure you have initialized your credentials via stagehand fn auth login using your Browserbase API keys.

Use Cases

• Automating routine data extraction from complex, JavaScript-heavy web applications.
• Building web-based agents that require authentication, such as logging into portals to download reports.
• Deploying long-running browser sessions for real-time monitoring or alerts.
• Creating webhooks that trigger browser-based actions based on external events.

Example Prompts

1. "OpenClaw, use the browse skill to help me create a script that navigates to the example.com login page and scrapes the daily price table into a JSON format."
2. "I need to automate filling out this form on a ticketing site every morning. How do I initialize a new project and configure the index.ts file for this task?"
3. "My deployment is failing with a 'No functions were built' error. Can you walk me through the mandatory package.json fixes required for my automation project?"

Tips & Limitations

• Always start by using the stagehand session create --local command to verify your logic before attempting to deploy to the cloud.
• The current CLI requires careful management of package dependencies; ensure your package.json strictly follows the pinned versions provided in the technical documentation to avoid build-time errors.
• Be aware that interacting with websites in this way is subject to the target site's Terms of Service; always implement polite crawling practices (e.g., timeouts, headers).
• Debugging can be intensive; utilize the snapshot and screenshot commands frequently to verify what the agent is 'seeing' at any given point in the DOM lifecycle.