Browser Interaction Safety

0•

Enforces strict security, privacy, and consent boundaries during AI-driven browser interactions and automated web tasks.

The Browser Interaction Safety skill provides robust guardrails for AI agents navigating and interacting with web interfaces. It dictates that browser interactions should only occur when strictly necessary for visual verification or dynamic rendering, favoring lightweight retrieval methods whenever possible. The skill protects sensitive information by prohibiting automated entry of passwords, payment credentials, or tokens, blocks screen captures containing secrets, and enforces stateless sessions to prevent cookie or credential leaks. Additionally, it treats all external web page content as untrusted to defend against prompt injection and requires explicit human consent before accepting legal agreements or terms of service.

Key Features

01Explicit human consent enforcement for terms of service, policies, and transactions
02Stateless navigation preventing persistent storage of cookies, sessions, and identifiers
03Capability-conditional browser activation prioritizing direct, non-browser data retrieval
04Anti-leak screenshot safeguards blocking captures of visible password or secret fields
05Strict credential protection preventing automated password entry and secret exposure
060 GitHub stars

Use Cases

01Navigating authenticated web workflows safely by delegating login and sensitive steps to the human user
02Researching and fact-checking information from untrusted third-party web pages without falling victim to prompt injections
03Safely automating UI verification and visual regression checks on dynamic web apps