Media
4 items
Description
Cast a sling ring, draw a circle, and step through a portal into any link. Draw a circle around a link. The page tears open inside the disc, the destination is already there, and you step through. SlingTab replaces the click with a gesture. You are reading an article, a link catches your eye, and instead of clicking you circle it. A rim of embers ignites along the exact path you traced, the page ruptures inside the ring, and through the hole you can see where you are about to go - live, before you commit. Then the well deepens, the hole swallows the viewport, and the navigation happens underneath. None of this is a theme or a skin. It is a different way of moving around the web, and it works on any ordinary web page. HOW YOU USE IT There are three ways to draw the circle. You pick one in the popup; only one is active at a time. 1. Right button. Hold the right mouse button and draw a circle around the link. The context menu is suppressed only when the gesture actually produced a portal, so you keep your right-click everywhere else. 2. Alt key. Hold Alt and move the pointer in a circle. No button needed, which suits a trackpad. 3. Your hand, in the air. Raise two fingers in front of your webcam and draw the circle. This one is off by default and has to be switched on deliberately - see the camera section below. Circle a link and the portal opens on that link. Circle empty space and, if you have configured quick links, they appear as chips around the rim: hover one to load it into the disc, then flick the pointer through the middle to go. A click outside the disc, or any key, closes it. Nothing is opened without you committing to it. WHAT YOU SEE INSIDE THE DISC SlingTab tries hardest to show you the real destination page, live, inside the portal. When the destination allows being framed, that is exactly what you get: the actual page, loading inside the disc while you decide. When it does not, or while it is still on its way, the disc shows a composed preview card built from the destination's own Open Graph image, title, description and theme colour, falling back to its favicon. Something correct is on screen the whole time, so a slow site never leaves you looking at an empty hole. The preview frame is scenery, not a browser. It cannot be clicked, it cannot submit forms, it cannot open pop-ups, and it cannot navigate your tab. It exists for the second or two the portal is open and is discarded when the portal closes. QUICK LINKS Up to eight destinations of your own, set in the options page. Paste a list - one link per line, address and name separated by a space - or add them one at a time. They are the chips that appear when you circle empty space, so a portal is never wasted on nothing. THE HAND TRIGGER, AND YOUR CAMERA This is off by default. If you never turn it on, SlingTab never asks for your camera and never opens it. If you do turn it on, the hand tracking runs entirely on your machine. The model - Google's MediaPipe Hand Landmarker - is bundled inside the extension rather than downloaded, so it works with no network at all. Camera frames are read, measured for a hand shape, and discarded. No video, no still image and no derived data ever leaves your computer, and none of it is written to disk. MediaPipe's own library ships a usage-telemetry client that batches events to a Google logging endpoint. SlingTab severs that endpoint at build time, and the build fails outright if the endpoint can no longer be found - so a future dependency upgrade cannot quietly reintroduce it. This is verifiable in the published source. Consent is collected on a dedicated extension page rather than in the popup, because Chrome closes a popup as soon as the permission bubble takes focus. That page also draws a live diagnostic view - the tracked hand, the finger pose, and the region of the camera frame that maps onto your screen - driven by the same tracker that fires the portal, so what you see there is what actually runs. PRIVACY SlingTab has no account, no server and no analytics. There is nothing to log into and nothing that phones home. What stays on your machine: - The snapshot of the page you are on. SlingTab captures the visible tab so the lens effect has something to bend. The image becomes a texture on your GPU and is dropped when the portal closes. It is never uploaded and never stored. - Camera frames, if you use the hand trigger. Read and discarded, as described above. - Your settings and quick links. Kept in Chrome's own extension storage, which syncs across your Chrome profile if you have Chrome sync switched on. Nobody else can read them. - Your browsing. SlingTab does not build a history, does not record which links you circle, and does not track you across sites. What does touch the network, and why: - The destination page you circled. To build the preview, the extension's service worker fetches that URL - the same URL your browser is about to open anyway - and reads the first 64 KB of it for meta tags. The request is made WITHOUT your cookies, so the site sees an anonymous visitor and no session of yours. - The destination's preview image, if it declares one, so it can be drawn inside the disc. Also fetched without cookies, and capped at 512 KB. - Favicons, which come from Chrome's own local favicon service and never leave the browser. Those requests go to the site you are travelling to, and to nobody else. There is no third party anywhere in the path. PERMISSIONS, AND WHY EACH ONE IS THERE storage - to save your trigger choice, your quick links and your preview preferences. Nothing else is kept. offscreen - to run the webcam in an offscreen document on the extension's own origin. Without it, getUserMedia would run in the context of whatever page you happen to be on, which would ask you for camera permission on every single site and attribute it to that site. The offscreen document means you are asked once, for SlingTab. It is only ever created while the hand trigger is on, and closed when it is turned off. favicon - to draw the destination's favicon on the preview card and on the quick-link chips. declarativeNetRequestWithHostAccess - this is the one that deserves a full explanation; see the next section. Note that it is the host-scoped form of the permission: it can only act on sites you have already granted access to, and it adds no install-time warning of its own. Host access to all sites - the gesture, the preview and the portal have to work on whatever page you are reading, and the page snapshot that drives the lens effect needs it too. SlingTab deliberately does NOT request the tabs permission: the service worker reads the tab id from the message sender, which needs no permission at all. It also requests no scripting, no webRequest, no cookies, no downloads, no history and no bookmarks. ABOUT THE FRAMING HEADERS Most sites send headers telling browsers not to display them inside a frame. Honouring those unconditionally would mean the portal almost never shows the real page, which is the entire point of the feature. So SlingTab can remove them - narrowly, briefly, and in a way that cannot be turned into an attack. Exactly what happens: - The rule is a SESSION rule, never a persistent one. It cannot survive a browser restart. - It is scoped to the single tab the gesture was made in. - It applies to sub-frame requests only, never to the top-level page you are on. - It targets only the hostname of the link you circled. - It is released as soon as the preview frame reports that it loaded; again when the navigation commits; again by a five-second watchdog; and again if the tab closes. On top of that, every rule the extension owns is swept away whenever the service worker starts, so one cannot be left behind by a worker that was shut down mid-gesture. - How much it removes depends on where the link goes. For a link to a DIFFERENT site it removes X-Frame-Options and Content-Security-Policy. That frame is cross-site, so the browser's own SameSite cookie rules withhold your session and the preview is of a logged-out page. For a link to the SAME site you are already on - where the frame would carry your cookies - it removes X-Frame-Options ONLY, and leaves the Content-Security-Policy, script-src included, completely intact. Why this is not a clickjacking surface: those headers exist to stop a hostile page from overlaying an invisible frame and stealing your clicks. SlingTab's frame cannot receive a click at all. It is rendered with pointer events disabled, inside a closed shadow root, sandboxed without form submission, without pop-ups and without top-level navigation, and with the referrer suppressed. A frame nobody can interact with is not a surface anybody can exploit. And if you would rather no header were ever touched, a switch in the options page turns this off completely. With it off, SlingTab frames only sites that already permit framing, and shows the composed card for the rest. NO REMOTE CODE Everything SlingTab runs is inside the package you install. No script is fetched at runtime, no CDN is contacted, no code is evaluated from a string, and no remote configuration is pulled. The hand-tracking model and its WebAssembly runtime are bundled, which accounts for most of the extension's size and is the reason it works with the network switched off entirely. PERFORMANCE With the default mouse trigger, SlingTab is idle until you make a gesture. There is no animation loop, no canvas and no timer running on the pages you visit - only event listeners. The WebGL context, the particle buffers and the page texture are created when a portal opens and released when it closes. The hand trigger is the expensive mode, and that is why it is not the default: it holds the camera open and runs hand inference continuously while it is selected. Turn it off and everything it was using is torn down. If your system is set to reduce motion, SlingTab skips the animation entirely and simply navigates. WHERE IT DOES NOT WORK On Chrome's own pages, the Web Store, the PDF viewer, and anywhere extensions are not permitted to run - every one of those cases degrades quietly to an ordinary navigation. Only http and https pages, and only at the top level: SlingTab does not inject itself into embedded frames. Requires Chrome 120 or newer, and a GPU capable of WebGL2 for the full effect. Where shaders will not compile it falls back to a plainer Canvas2D rendering rather than failing. OPEN SOURCE SlingTab is MIT licensed and the complete source is public, including the build configuration that severs the telemetry endpoint and the tests that pin the behaviour described above. https://github.com/AtlasAnatomy/SlingTab Bundled third-party components - Google's MediaPipe Tasks Vision under Apache 2.0, and the Alumni Sans and Albert Sans typefaces under the SIL Open Font License - are attributed in full, with their licence texts, inside the extension package itself.