A CLI flag is not a security boundary, and one browser tool just proved it

Description and use

<p>Tencent released BrowserSkill, a tool that lets AI agents drive the browser you are already logged into. The obvious story is the capability. The interesting story is a breaking change buried in version 0.3.0.</p> <p>Before 0.3.0 you could run an agent's browser session unattended by passing <code>--unattended</code>, or <code>tab borrow --no-confirm</code>, or setting <code>BSK_REQUEST_HELP=off</code>. Each one skipped the confirmation prompt that normally asks before the agent takes control of one of your tabs.</p> <p>In 0.3.0, none of them work any more. They are still accepted, they are

Problem

<p>Tencent released BrowserSkill, a tool that lets AI agents drive the browser you are already logged into. The obvious story is the capability. The interesting story is a breaking change buried in version 0.3.0.</p> <p>Before 0.3.0 you could run an agent's browser session unattended by passing <code>--unattended</code>, or <code>tab borrow --no-confirm</code>, or setting <code>BSK_REQUEST_HELP=off</code>. Each one skipped the confirmation prompt that normally asks before the agent takes control of one of your tabs.</p> <p>In 0.3.0, none of them work any more. They are still accepted, they are

Outcome

<p>Tencent released BrowserSkill, a tool that lets AI agents drive the browser you are already logged into. The obvious story is the capability. The interesting story is a breaking change buried in version 0.3.0.</p> <p>Before 0.3.0 you could run an agent's browser session unattended by passing <code>--unattended</code>, or <code>tab borrow --no-confirm</code>, or setting <code>BSK_REQUEST_HELP=off</code>. Each one skipped the confirmation prompt that normally asks before the agent takes control of one of your tabs.</p> <p>In 0.3.0, none of them work any more. They are still accepted, they are

Evidence

A CLI flag is not a security boundary, and one browser tool just proved it

Source