The Agent Skills Directory

[SAFE]: The skill incorporates a security-conscious workflow by mandating the use of AskUserQuestion to obtain user consent before proceeding with code generation or modifications based on recorded browser sequences.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes functionality for capturing and managing sensitive session data, such as cookies and localStorage (browser_get_cookies, browser_export_session). These tools are used appropriately for the stated purpose of maintaining authentication in automated workflows.
[DYNAMIC_EXECUTION]: JavaScript execution is supported through RunScript nodes to facilitate DOM manipulation and data extraction. This is standard functionality for browser automation tools and is documented with clear guidance for the agent.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external web content, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through browser_navigate and browser_snapshot operations.
Boundary markers: No specific delimiters or warnings for the agent to ignore instructions within web content are provided in the skill instructions.
Capability inventory: The skill provides access to browser actions, network monitoring, and local file writing for session persistence.
Sanitization: The skill does not define specific sanitization or filtering protocols for data ingested from the browser before it is processed by the agent.

exploring-browser