The Agent Skills Directory

Execution of JavaScript in Browser Context: The eval command allows the agent to execute JavaScript code directly within the browser's environment. While this is a standard feature for advanced automation and data extraction, it involves the dynamic execution of scripts in the browser sandbox.
Authentication and Session State Management: The skill supports saving and loading browser state (e.g., state save auth.json). This process involves handling sensitive session data like cookies and local storage to maintain authenticated sessions across tasks.
External Data Processing and Interaction: The skill is designed to interact with and extract information from external websites, which introduces a surface for indirect prompt injection.
Ingestion points: External web content is ingested via agent-browser open <url>, snapshot -i, and get text (SKILL.md).
Boundary markers: The skill does not define specific delimiters or warnings to ignore instructions that might be embedded in web pages (SKILL.md).
Capability inventory: The tool possesses high-interaction capabilities, including click, fill, eval, state save, and press (SKILL.md).
Sanitization: There is no mention of sanitizing or filtering external content before it is processed by the agent (SKILL.md).
File System Operations: The skill performs file writes to store automation artifacts, such as screenshots, PDFs, and extracted page text. These operations facilitate debugging and reporting of the automation tasks.

agent-browser