The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes browser-use python "code" and browser-use python --file script.py commands, which allow for the execution of arbitrary Python code within the agent's environment.
[REMOTE_CODE_EXECUTION]: The browser-use eval "js code" command enables the execution of arbitrary JavaScript code within the browser context.
[DATA_EXFILTRATION]: The browser-use tunnel <port> command allows exposing local network ports to the public internet via Cloudflare tunnels, which can be misused to exfiltrate local data or expose internal services.
[DATA_EXFILTRATION]: The skill facilitates data collection through browser-use screenshot, browser-use cookies export, and browser-use get html commands.
[DATA_EXFILTRATION]: The browser-use upload <index> <path> command allows the agent to upload local files to external websites.
[CREDENTIALS_UNSAFE]: The browser-use cookies get and browser-use cookies export commands allow for the extraction of browser session cookies, which often contain sensitive authentication tokens.
[COMMAND_EXECUTION]: The skill relies on the browser-use CLI tool for all browser automation and system-level interactions.
[PROMPT_INJECTION]: The skill creates a significant indirect prompt injection surface by ingesting untrusted web data while maintaining high-privilege execution capabilities.
Ingestion points: Untrusted data enters the agent context via browser-use state, browser-use screenshot, and browser-use get html (SKILL.md).
Boundary markers: There are no mentioned delimiters or instructions to ignore embedded commands in the ingested web content.
Capability inventory: The skill provides access to Python execution, JavaScript evaluation, network tunneling, and file uploads (SKILL.md).
Sanitization: No sanitization or content filtering mechanisms are described for the data retrieved from external websites.

remote-browser