* fix(studio): recovery drafts are kept by each project's own server, never the browser
* fix(studio): a restored draft stays offered when Studio settles which composition is open
* fix(studio): a draft read never waits on a pipe or loads an oversized record
* test(studio): the draft read's child fixture is an entry point, and its note sits in the test
* chore(studio-server): size budget covers the recovery-draft route
* chore(cli): unpacked size budget covers the recovery-draft route
* fix(cli): the CLI trusts a preview only once it proves this user started it
* test(cli): a listener that cannot prove it is the user's preview is never trusted
* test(cli): stand-in previews answer the challenge as the user's own do
* style(cli): format the preview proof changes
* fix(cli): a preview's proof names its own endpoint and config, and the token file must be private
* style(cli): format the proof changes
* test(studio): the fake request in the host check has a socket, as every real one does
* fix(studio): a Studio started by hand proves it is the user's, so the CLI still finds it
* fix(cli): the challenge travels in a header, so Studio's routes keep their exact paths
* style(studio): keep the config producer compact
A preview made opaque by its server (Content-Security-Policy: sandbox) under a
frame with no sandbox attribute posts as null. frameOrigins now accepts null
for such a frame, only from that frame's own window and only while its window
refuses reads.
* fix(studio): the preview reads hf-preview messages only from its own frame
The preview message router acted on a message with no sender window, and on
any message that arrived before the preview frame existed. It now takes an
hf-preview message only when the preview frame exists, the message comes from
that frame's window, and its origin is one the frame can have: the page's own
origin, or "null" for a sealed frame. The element picker checks the same.
* fix(cli): present sandboxes deck pages another site loads and refuses framing by other sites
present now applies the same another-site decision as play to its
/composition/* route: a load by another site gets a sandbox CSP, its
validators are dropped, and the document is marked so its runtime offers
that site no editing port. Every present response also sends
frame-ancestors 'self' and X-Frame-Options SAMEORIGIN, so only the
presenter page itself can frame the deck. The route setup moves into
buildPresentApp so tests exercise the server as run() builds it.
* chore: release v0.9.55
* chore: release v0.9.55 notes include #7830
* chore: release v0.9.55 notes include #7859 and #7867
* chore: release v0.9.55 notes name the docs change and cloud render
* chore: release v0.9.55 notes include #7866 and #7801
* chore: release v0.9.55 notes match the style bar
* feat(studio): hosts can seal the preview frames NLEPreview draws
* refactor(studio): Player builds its frame src in a helper, keeping its complexity at main's
* refactor(core): Studio's key combo and typing-target rules live in core
* chore(core): raise the packed size budget for the key-combo and typing-target subpaths
* fix(deps): bump @mcp-b/global to 6.1.0 to clear @modelcontextprotocol/server advisories
@mcp-b/global 5.x and 6.0.x pin @modelcontextprotocol/server 2.0.0, which is
affected by GHSA-f897-62hf-6gxq, GHSA-46jr-4wx8-v453, GHSA-jm63-cq47-j5jp and
GHSA-rvq5-wwqv-78pq (>=2.0.0 <2.3.0). 6.1.0 is the first release that pins
server 2.4.0.
6.x installs the W3C WebMCP polyfill and passes { signal } to execute; the
real-package tests now drive the 6.x surface (secure context, MCP tools/call).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(studio): close the WebMCP polyfill after the real-package suite
6.x's BrowserMcpServer sends a tools-changed event on a timer after the
last tool unregisters. Without cleanupWebModelContext() it fired after jsdom
teardown and vitest failed the run on an unhandled rejection. Also scope the
transport test's header to the MCP transports it actually checks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(studio): menu row highlight follows the menu's corner
* fix(studio): keep the menu row's comment share flat
* fix(studio): export the menu row class so hosts reuse it
* feat(studio): press ? to open the keyboard shortcuts panel
Adds a rebindable `shortcuts` action bound to `?` (Shift+/) that toggles
the shortcuts panel; Esc still closes it. The key is ignored while focus
is in an input, textarea or contenteditable, and the panel lists `?`
under Panels.
Refs heygen-com/hyperframes#5580
* docs(studio): document the ? shortcut for the shortcuts panel
* fix(studio): edit clips added to a project outside Studio
A clip written into a composition after its ids were last stamped (a hand
edit, an outside agent) has a data-hf-id only in the served preview. The
timeline's client-side patch looked that id up in the file's raw bytes, so
Hide, Mute, move, resize and attribute saves failed with "Unable to patch
timeline element ... (by hf-id; no file on the clip)".
Resolve the target against the file stamped in place the way the preview
stamps it, as the server's own saves and the GSAP save already do. The save
then persists that id.
* fix(studio): save volume on several clips added outside Studio
The multi-clip attribute save (gain or volume on a selection) patched each
clip in the file's raw bytes, so a clip with no ids on disk was skipped and
the save reported success with nothing written.
The by-target read and patch helpers now resolve a clip by the id the
preview served it with, so every caller finds it. The multi-clip save goes
through the same batch patch as the other timeline saves, which refuses a
clip the file does not hold; the error reaches the person instead of a
silent no-op. The preview picker finds the file holding an unstamped
element the same way.
* feat(cli): check warns when text leaves the 9:16 social safe areas
* feat(core): safe areas owned by core; check adds --safe-zone standard
* style(cli): format
* refactor(core): name the safe-area band fields
* fix(cli): defects outrank safe-zone warnings at the issue cap
* style(cli): format
* test: read the safe-box constants from core
* feat(studio): drag and resize on an opaque preview start from one gesture-start read
A gesture on a preview frame Studio cannot script read nothing from the
frame as it started. A drag snapped to nothing; a layer GSAP holds moved by
a CSS translate that its save then dropped; a resize grew from the overlay
rect, the layer's GSAP scale included; and the pointer reached the element
only through the canvas zoom.
The runtime answers a new `gesture-start` request. For each pressed element
it gives the overlay geometry, the transform-free layout size, how far the
centre moves per px of translate (probed the way a draft draws it, every
transform above it applied), and GSAP's x/y when GSAP owns its position,
with whether a tween moves it. It also lists the boxes of the page the
gesture can snap to, walked by the rule Studio's own snap collection now
takes from core.
On an opaque preview every press waits for that one read, on the existing
waiting-press path, and the gesture starts from it: the snap targets, the
move route and GSAP base (so the save writes GSAP's x plus the drag, not the
drag alone), the pointer-to-offset matrix, and the resize's base size. A
layer a tween moves is refused with a notice, since saving that move still
reads the frame.
Same-origin frames do not change: their presses start at once from the
element, and their snap collection walks the same elements by the same rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore(inspector): raise the packed size budget for the gesture-start runtime
The inspector bundles core's preview runtime. Main alone packs to 866206
bytes, 420 under the 1% tolerance on the old 858046 budget; the
gesture-start request adds 713 bytes (866919), which tipped the merge
queue's Build over. Set the budget to 880000 to leave room for the
remaining opaque-preview requests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
oxfmt 0.41 pins tinypool 2.1.0 (GHSA-5gmw-xhrv-c9v3, GHSA-85c8-ppgw-ccpr).
Every oxfmt release before 0.67 pins 2.1.0, and 0.67+ reformats 66 files,
so override tinypool to the drop-in 2.1.2 patch instead of bumping oxfmt.
Also bump vite ^6.4.2 -> ^6.4.5 and vitest/@vitest/coverage-v8
^4.1.11 -> ^4.1.12 (pulls @vitest/mocker 4.1.12).
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(studio): drag and resize draw live on an opaque preview over its channel
On a preview frame Studio cannot script, a drag moved only the selection
outline: the element stayed put until the save reloaded the preview. A
resize threw on every pointer move (it measured the frame's element for the
centre pin), saved its size without the centring shift, and then a drag or
resize save threw while re-syncing the preview's timeline, which stopped the
save queue and paused the preview.
The runtime answers a new `draft` request: it adds a gesture's translate to
the one the element had and sets its border box, the way the save writes
them, and an end either keeps what is drawn (the save landed) or puts back
what was there (a cancel, a tap, a failed save). Studio mirrors what a
gesture draws on an opaque selection's stand-in to the frame, one request in
flight per frame with later moves merged into the next. An opaque resize
grows the outline about its centre from the size it wrote rather than
measuring the frame. The timeline re-sync after a save reports a frame it
cannot reach, so its caller re-seeks or reloads instead of throwing.
A stand-in now starts with the inline translate and Studio size and offset
of the element it stands for, so a drag on a freshly reloaded preview starts
from where the element is. Before, a second drag after a reload saved only
its own distance and dropped the first.
Same-origin frames do not change: only an opaque selection's stand-in is
mirrored, and the re-sync change only catches a read such a frame refuses.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(core): draft writes test an element by its namespace, not its realm
The runtime preview guard bans instanceof HTMLElement/SVGElement in the
runtime, since a node from another realm fails it; the draft handler uses
isStylableElement from domRealm.ts, as other runtime style writes do.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(studio): keep the opaque preview's gesture drafts in step with the saved edit
- Grabbing a layer again before its last save settles keeps that save's draft
on the frame and draws the new gesture from there; the movement probe's
restore no longer ends a draft it never drew.
- A move on an opaque preview starts from the px the frame renders the
translate at (%, em, var(), or a stylesheet translate), records the author's
own value for Reset, and refuses a % it cannot size instead of dropping it.
The frame's draft base resolves % to px, so a resize can't shift it again.
- The stand-in carries Studio's translate marks, so Reset after a reload
restores the author's translate.
- A resize with no offset member keeps its draft once saved.
- An undo painted over a pending move draws the move again when it is put back.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* chore(deps): update happy-dom to ^21.0.4
happy-dom 21 turns off iframe page loading by default. The player and
sdk test environments load same-origin iframe documents, so they opt
back in through their vitest happyDOM settings.
— Clippy
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test: adapt studio and cli tests to happy-dom 21
Studio tests load same-origin iframe documents, so they opt back in to
iframe page loading. happy-dom 21 adds Element.checkVisibility, which
reads the real styles rather than the mocked getComputedStyle, so the
motion-signature decoys are hidden in the DOM as well as in the mock.
— Clippy
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(studio): fix remaining studio tests on happy-dom 21
- happy-dom now implements getAnimations, so Base UI deferred popup
unmounts to animation frames; test setup sets Base UI's
BASE_UI_ANIMATIONS_DISABLED, matching the old synchronous behaviour.
- happy-dom 21 drops vendor-prefixed properties it does not list, so the
-webkit-text-fill-color tests move to a jsdom file with shared helpers.
- The scaled-iframe drag test builds its own Window, which now needs
enableIframePageLoading.
- NLEPreview.slowBoot's per-file environment options never applied the
disable setting and now wiped the config's iframe opt-in; drop them.
— Clippy
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(studio): inline fill-test helpers instead of a shared module
Keeps the moved -webkit-text-fill-color tests self-contained and leaves
the original test file's helpers untouched.
— Clippy
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(studio): drop the Base UI flag comment for the comment ratchet
— Clippy
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(studio): a file held open or read-only answers 423 or 403 with a reason, not 404 or 500
* fix(studio): valid workflow name for the Windows held-file test
* test(studio): release the held file before the held-file test removes its folder
* fix(studio): the code editor's save toast names a held or read-only file
A pointer move superseded by a newer move or a leave returned null, not just
skipped the hover update. The right-click path awaits that same lookup for its
target when nothing is selected, so on a sandboxed preview, where the lookup
crosses the frame's channel, a 1px jitter during the right-click bumped the
counter and the menu never opened. The stale hover is still dropped; the
result is returned either way.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(studio): layers panel and marquee read an opaque preview over its channel
The runtime answers a new layers request with the Layers walk: each layer's
handle, depth, child count, z-index and kind, by the rules Studio walks a
scriptable frame with, which move to core. On an opaque frame the Layers
panel lists, selects and drills in from it, and the marquee measures its
candidates over the channel and selects from their descriptions. An opaque
selection maps to a timeline clip known only by its clip id, so a marquee's
picks survive the timeline sync.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(studio): opaque preview hover, layers drill-in, marquee paging and delete path
Review follow-ups for the opaque-preview Layers panel and marquee:
- Delete on an opaque selection runs the DOM delete again. The clip-id match
that lets a selection find a row known only by its clip id is now opt-in
(`clipIdScope`), passed only by selection sync and the Layers seek, never by
delete or group commits, which need a patch target that row lacks. A row
with no file reads as in the open composition, as the forward lookup does.
- The runtime's `layers` reports a drill-in group it no longer has
(`missingGroup`) and lists nothing, so the panel leaves the drill-in after a
reload as the same-origin branch does, rather than list the whole page.
- `layers` takes an `offset`, and the marquee pages through every layer, so
elements past one reply's 2,048 cap still band-select.
- A Layers refresh that lands while an opaque read is in flight runs once more
when it settles, instead of being dropped.
- A hover resolved for a pointer move that a leave or newer move superseded is
dropped.
- `pointer-targets` reads the hit stack once per request.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(core): keep the layers paging test inside the Windows runner's timeout
The paging test walked the 2,051-layer page three times with every row
mapped, which ran past 5 s on windows-latest. It now maps only the last
page's rows and has its own 20 s timeout.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(studio): canvas click, hover and timeline select read an opaque preview over its channel
The runtime answers a new pointer-targets request with the canvas pick and its
cycle stack; the hit-test moves to core so Studio picks by the same rule on a
frame it can script. A data: value under the description limit now arrives
whole, so a small data: image is selectable on an opaque frame. Drill-in,
hover, timeline clips known only by clip id, and the stand-in attribute reads
in double-click, ungroup, path-offset and the agent selection context work
there too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore(core): raise the core unpacked size budget for the pointer-targets hit-test
The hit-test moves from Studio into core's editing runtime, which core ships
in its runtime bundles; CI measured the unpacked tarball at 4,329,030 bytes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(studio): read overlay geometry from an opaque preview over its channel
The runtime's editing channel gains a `geometry` request: per element
handle its client rect, the corners it paints at when rotated, whether it
renders, its clip-path, its source boundary's size, a group wrapper's
member rects, and optionally one element's descendant boxes. On a preview
frame Studio cannot script, the selection, child, group and hover boxes,
the root scale (from the mirrored stage size) and a press held across a
preview change read that geometry, with one read in flight at a time.
Same-origin frames keep reading the DOM directly; the measuring code they
use now lives in core so both paths share one rule.
Also from the review of #7613: opaque selections carry `hasAudio` and
`src` read off the description, a description with a cut attribute or
inline style builds no selection, a failed computed-styles or root read
builds none either, and the group refresh threads its frame and leaves a
group the user changed meanwhile. `findClosestByAttribute`, left without
a caller by #7613, is removed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore(producer): raise the packed size budget for the runtime's geometry read
The producer ships the preview runtime, which now carries the editing
channel's `geometry` request and the measuring code it shares with Studio.
The packed tarball measured 888,810 bytes in CI against an 880,000 budget.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(studio): timeline select and selection refresh read an opaque preview over its channel
On a preview frame Studio cannot script, timeline select, a selection
refresh and a group refresh build the selection from the runtime's element
descriptions. The rules that pick the element (group capture and drill-in,
the clip ancestor, the walk to an addressable element, timeline-locked, the
lookups and the master-view root fallback) move to a pure module over an
ancestor chain, used by both paths, so both select the same element.
Same-origin frames keep their direct reads.
The editing port gains `ancestors` (each handle's chain, parent first) and
`query` (a selector's matches with their source facts), and
`computed-styles` also names its element by handle. An opaque selection
carries the element's handle; its `element` is a tag-only stand-in in a
document with no window.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore(studio): keep the selection rules' comment share at the package's
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>