Yes, in a sandbox. litehtml itself never runs scripts; a message's <script> runs in its own JavaScriptCore context with no network and no DOM. All it can do is return HTML from render() and react to clicks in onAction.

Every question has data-action="open". A click calls onAction('open', id); the script keeps a single openId, so opening one answer closes the other, calls transition('slide-down'), and render() returns the new HTML for this same bubble.

Never. Each bubble is a separate litehtml document with its own stylesheet list, so a <style>p { color: red }</style> only affects its own bubble.

There is no hard limit. The article above carries around 20 KB of HTML; the leaderboard in “Analytics” has fifty table rows. Layout cost grows roughly linearly with the number of words and boxes, and a script may return up to 256 KB per render.

Yes — paste HTML into the composer, with <style> and <script> if you like, or edit the .chat files in Resources/chats. Anything you send shows fullsize; wrap it in <div class="widget"> to show it inside a bubble.

Every call into a script is capped at 50 ms, and a message gets at most 200 ms of CPU per second. Break either rule and the script is stopped; the card keeps its last HTML and offers a Restart link.

No transitions, transforms, opacity, box-shadow or grid. Scripts animate with what is left — rgba() colours, margins, sizes and absolute positions — re-rendering up to 60 times a second.

No. Timers and frames pause while a bubble is off screen and catch up when it comes back, so a long chat full of live cards costs nothing.