You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Desktop: user CSS can reveal normally hidden Host bootstrap scripts in head (repro and tested containment)
#12680
Independent Host-side reproduction (not a diagnosis of a user's screenshot)
In an isolated Windows Node 24.21 / official DSH 0.2.2-rc.1 Electron Desktop profile, third-party/user CSS that resets native element display can expose the normally hidden <head> and render DSH's own inline bootstrap JS/CSS visibly at the top of the Desktop screen, despite document.contentType === "text/html". This can visually match a reported "raw JavaScript/CSS wall" without a plugin response-MIME or parser failure.
No Vision Router required: an independent clean-profile A/B/A with only the Host and pinned external dsh-custom-style CSS plugin reproduced the raw-source display when a deliberately synthetic * { all: unset !important } rule was loaded; removing that one CSS rule restored the hidden head in the same Electron process. Initial cold-launch reproduction was also confirmed.
Tested containment candidate
The existing Desktop document owns the Host bootstrap <head> and its inline __DSH_BOOT_READY__. In isolated Host source only, making the generated <head style="display: none !important"> explicitly nonrendering prevented the boot source becoming visible while the hostile CSS remained loaded. The Host boot promise and the normal Vision image→tool→chat flow continued to work. A real Chromium matrix covering ordinary, universal reset/display and explicit head styles passed 8/8, and a focused Vitest test was red on original upstream code / green on the candidate.
This proves a general Desktop HTML/CSS containment issue; it does not establish that the original community reporter actually installed this CSS or identify their true cause.
Global CSS resets may also damage other UI elements; protecting the head is not a universal theme/plugin fix.
No user profile, session data or release artifacts were changed in these probes.
Upstream Issues are disabled and a fork-to-upstream PR attempt under this contributor account returned CreatePullRequest permission denied. Could the maintainers review the minimal Host-owned containment and advise the preferred protected head/document behavior?
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
Independent Host-side reproduction (not a diagnosis of a user's screenshot)
In an isolated Windows Node 24.21 / official DSH 0.2.2-rc.1 Electron Desktop profile, third-party/user CSS that resets native element display can expose the normally hidden
<head>and render DSH's own inline bootstrap JS/CSS visibly at the top of the Desktop screen, despitedocument.contentType === "text/html". This can visually match a reported "raw JavaScript/CSS wall" without a plugin response-MIME or parser failure.No Vision Router required: an independent clean-profile A/B/A with only the Host and pinned external
dsh-custom-styleCSS plugin reproduced the raw-source display when a deliberately synthetic* { all: unset !important }rule was loaded; removing that one CSS rule restored the hidden head in the same Electron process. Initial cold-launch reproduction was also confirmed.Tested containment candidate
The existing Desktop document owns the Host bootstrap
<head>and its inline__DSH_BOOT_READY__. In isolated Host source only, making the generated<head style="display: none !important">explicitly nonrendering prevented the boot source becoming visible while the hostile CSS remained loaded. The Host boot promise and the normal Vision image→tool→chat flow continued to work. A real Chromium matrix covering ordinary, universal reset/display and explicit head styles passed 8/8, and a focused Vitest test was red on original upstream code / green on the candidate.Reviewable fork-only candidate: ysr666@06b8896 . Detailed DVR investigation: ysr666/dsh-vision-router#792 .
Boundaries / ask
CreatePullRequestpermission denied. Could the maintainers review the minimal Host-owned containment and advise the preferred protected head/document behavior?All reactions