The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the browseros-cli tool to execute browser actions such as navigation, element interaction, and data extraction through the shell environment.
[EXTERNAL_DOWNLOADS]: The instructions guide the user to download the CLI via npm and the BrowserOS application using browseros-cli install from vendor-managed infrastructure.
[DYNAMIC_EXECUTION]: The skill exposes the eval command, which allows for the execution of arbitrary JavaScript within a browser tab, representing a powerful dynamic code execution capability.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted external data from websites, creating a surface for injection attacks.
Ingestion points: External content is brought into the agent context via snap, text, dom, and links commands (defined in SKILL.md and references/cli-commands.md).
Boundary markers: There are no explicit instructions for using delimiters or boundary markers to distinguish untrusted web data from system instructions.
Capability inventory: The skill has access to high-impact capabilities including eval (JS execution), form manipulation (fill, click), file handling (upload, download), and access to sensitive browser resources like history and bookmarks.
Sanitization: No sanitization or validation steps are documented for the content extracted from web pages before it is analyzed by the agent.

browseros