The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to interface with the operating system and execute the browse CLI for browser automation, cloud API management, and skill installation.
[EXTERNAL_DOWNLOADS]: The skill includes instructions to install the browse CLI package globally using npm. Additionally, it features a mechanism to discover and install site-specific automation skills from the Browse.sh catalog using browse skills add.
[INDIRECT_PROMPT_INJECTION]: The skill provides an extensive surface for indirect prompt injection as it is designed to ingest and process untrusted content from the web.
Ingestion points: Functions such as browse open, browse snapshot, browse get markdown, browse cloud fetch, and browse cloud search bring external web content into the agent's context.
Boundary markers: The instructions do not define specific delimiters or "ignore previous instructions" markers for the content retrieved from the web.
Capability inventory: The agent possesses powerful capabilities including bash command execution, browse eval for JavaScript execution, and browse upload for uploading local files to remote sites.
Sanitization: There are no built-in sanitization or validation steps mentioned to filter potentially malicious instructions embedded in the ingested web data.
[DYNAMIC_EXECUTION]: The skill supports the execution of dynamically provided or generated code through several channels:
The browse eval command allows for the execution of arbitrary JavaScript within the active browser page.
The browse functions suite allows the agent to develop, publish, and invoke serverless functions on the Browserbase platform.

browse