The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the installation and maintenance of the agent-browser CLI and its browser engines from remote sources.\n
Documentation describes installation via npm, brew, and cargo.\n
The agent-browser install and upgrade commands are used to download and update the Chromium browser engine and the tool itself.\n- [DYNAMIC_EXECUTION]: The skill includes an eval command that allows the execution of arbitrary JavaScript within the browser context.\n
The functionality supports execution via simple expressions, Base64-encoded strings, or standard input to manage complex scripts and bypass shell escaping constraints.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the web, which creates a vulnerability surface for indirect prompt injection attacks.\n
Ingestion points: Data is brought into the agent's context through snapshot, get text, and screenshot operations across multiple files.\n
Boundary markers: The documentation recommends using the AGENT_BROWSER_CONTENT_BOUNDARIES environment variable to wrap external content in nonce-based delimiters.\n
Capability inventory: The tool provides extensive interaction capabilities, including element manipulation, network request routing, and clipboard access.\n
Sanitization: The platform provides opt-in content boundaries and domain allowlisting (AGENT_BROWSER_ALLOWED_DOMAINS) to restrict the agent's exposure to potentially malicious content.

agent-browser