The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the @browserbasehq/browse-cli package from the npm registry. This package is provided by a well-known service for browser automation.
[DYNAMIC_EXECUTION]: The browse eval command, described in REFERENCE.md, allows the agent to execute arbitrary JavaScript code within the context of the active web page.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from arbitrary external websites, which serves as a potential vector for instructions hidden in web data to influence agent behavior.
Ingestion points: Data from external URLs is brought into the agent's context through navigation and snapshot commands as defined in SKILL.md and REFERENCE.md.
Boundary markers: There are no instructions for using delimiters or specific prompts to differentiate untrusted web content from the agent's primary instructions.
Capability inventory: The skill permits a wide range of browser interactions, including clicking, typing, and executing JavaScript, which are performed via shell commands.
Sanitization: The skill does not implement or mention any sanitization or validation of the data retrieved from the browser before it is processed.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via the Bash tool to interact with the browse CLI utility for all browser operations.

browser