The Agent Skills Directory

JavaScript Execution: The skill utilizes an eval command to execute JavaScript within the browser context. This is a standard and necessary feature for automation tools to interact with complex web pages, though it represents a dynamic execution surface within the browser environment.
Session Management: The skill provides commands to save and load browser state, such as authentication cookies. While useful for maintaining authenticated sessions across tasks, it involves handling sensitive session data that users should ensure is stored securely on the filesystem.
Web Content Interaction (Indirect Prompt Injection): The skill is designed to navigate and process data from external websites, which creates a potential surface for indirect prompt injection where malicious content on a webpage might attempt to influence the agent. This is an inherent consideration for any tool processing untrusted data from the internet.
Vulnerability Surface Analysis (Indirect Prompt Injection): (1) Ingestion points: Untrusted data enters the agent context through the open and snapshot commands in SKILL.md when visiting external URLs. (2) Capability inventory: The tool can perform file writes (screenshots, state files), navigate the network, and execute actions within the browser as described in SKILL.md. (3) Boundary markers: The instructions do not define specific delimiters to separate untrusted web content from agent instructions. (4) Sanitization: The skill does not specify explicit sanitization steps for data retrieved from web pages before it is processed by the agent.

agent-browser