feat(registry): curtain-reveal transition

[miguel-heygen]

What

Adds curtain-reveal to the catalog: two pleated stage curtains that gather outward from center, the way real velvet bunches at the tieback, revealing the stage underneath. A static pelmet frames the top; a soft spotlight can bloom in as the curtains part.

Why

"Curtain" returns zero results on the catalog's word-tier search today; agents asking for a physical theatre-curtain reveal have nothing to point at.

Look

Each curtain is a pleated panel (a repeating linear-gradient standing in for fabric folds) with a closed-form sine-wave scalloped hem sampled at a fixed point count, clipped via clip-path. Both curtains part on one authored GSAP tween each with explicit fromTo endpoints; an optional pelmet (its own scalloped trim) stays static, and an optional spotlight glow blooms in as they part.

Structure

Copies the catalog item layout from registry/components/ink-bleed-reveal/ — slot-mount pattern at registry/components/ink-bleed-reveal/ink-bleed-reveal.html:233-247, registry-item.json shape, docs page structure from docs/catalog/components/grade-split-reveal.mdx.

Determinism

No physics state: the scalloped hem is a closed-form sine sample, both curtain tweens have explicit endpoints, no Math.random/wall clock/requestAnimationFrame anywhere. The spotlight sub-tween is clamped so it always ends at or before the curtain-open envelope, so a very short composition (below the base open duration, with no hold) never runs past its declared data-duration.

Rendered demo capture is attached above (rendered via devbox headless Chrome since ai-miga was mid-outage team-wide).

Test plan

• comment-check clean against origin/main
• Independent adversarial review, no findings above the confidence bar (report on file)
• Rendered demo capture attached
• Seen by home and Miguel before merge

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
hyperframes	🟢 Ready	View Preview	Sep 23, 2026, 11:30 AM

💡 Tip: Enable Automations to automatically generate PRs for you.

[miguel-heygen]

Rendered demo (rendered via devbox headless Chrome since ai-miga was down for this branch).

curtain-reveal-fixed.mp4

[terencecho]

Reviewed at 548de3fd.

Scope. 11 files: registry/components/curtain-reveal/{curtain-reveal.html, demo.html, registry-item.json} + docs/catalog/components/curtain-reveal.mdx + entries in docs.json, catalog-index.json, local-vectors.json/bin, catalog-gallery-data.mdx, and the new preview JSON blob. Additive; no touch outside the curtain-reveal item.

Component correctness (curtain-reveal.html).

• Deterministic. No Math.random, no wall clock, no requestAnimationFrame, no Date.now. Pleat texture is a static repeating-linear-gradient. Scalloped hem via hemPolygon(scallops, ampPct, baseY) samples sin(t * PI * scallops) * ampPct at 25 fixed points (STEPS = 24, i inclusive of both ends) — same trim every render.
• Envelope clamped to data-duration. IN = Math.min(IN_BASE, duration) where duration = Math.max(0.001, parseFloat(root.dataset.duration || "2.4")), so the fixed-IN case reduces to a fully-compressed part when data-duration < IN_BASE. The two curtain fromTo tweens have duration: IN and start at 0, so they land at exactly IN.
• Spotlight sub-tween mathematically pinned to end at IN, not past it. spotlightStart = IN * 0.25, spotlightDuration = Math.min(IN * 0.85, IN - spotlightStart). IN - spotlightStart = 0.75 * IN, min(0.85 * IN, 0.75 * IN) = 0.75 * IN, so the spotlight ends at spotlightStart + spotlightDuration = 0.25 * IN + 0.75 * IN = IN exactly. The uncapped version (duration = IN * 0.85) would land at IN * 1.10 and, on a data-duration < IN_BASE mount, escape past data-duration. The clamp is right and the comment above it names the exact failure mode it defuses.
• tl.set({}, {}, duration) anchors timeline length so GSAP's auto-extend doesn't push past data-duration.
• Tone table is safe. Object.prototype.hasOwnProperty.call(toneTable, vars.tone) guards against prototype pollution — a vars.tone = "toString" won't resolve toneTable[tone] to a stringified function. Falls back to "red" for unknown values.
• Slot pattern. Uses hostDoc.querySelector('template[data-slot="curtain-reveal-stage"]') at document level (templates never render in-place, so the caller's markup is inert until this primitive imports it). Falls back to the built-in .cvr-stage-default card if no slot template is present. defaultMark.remove() before appendChild(hostDoc.importNode(...)) — no double-render.
• Mount contract. Registers one paused timeline under literal "curtain-reveal" key. #root is position: absolute; inset: 0; container-type: size; isolation: isolate; — CSS cqw / cqmin units size correctly against the host box, no data-width / data-height required.
• No listeners, timers, observers → no teardown needed beyond the runtime's element removal.

Demo (demo.html). Standalone composition mounting the primitive full-bleed via data-composition-src, navy tone with pelmet off — exercises the non-default variable path. Slots a procedural inline-SVG stage scene so the demo ships nothing licensed (matches the grade-split-reveal pattern). Correctly reaches into the template's content via template.content.getElementById(...) rather than document.getElementById (inert templates aren't in the live document).

Catalog-drift check. Cleaner scope than #4331 — docs.json only bumps the Motion & effects (94 → 95) and Effects (53 → 54) tag counts and inserts catalog/components/curtain-reveal alphabetically between confetti and cut-the-curve. catalog-index.json inserts a single entry alphabetically between cursor-glyph-trail and cut-the-curve. local-vectors.json inserts curtain-reveal alphabetically. No unrelated drift folded in. catalogArtifact.revision bumps to 2c3b540a….

Registry.json placement. curtain-reveal appended at the end of items[] after warm-grain — note this is different from #4331's shape (which reorders the 8 hyperframes:example items to the top). Both PRs bump the artifact revision from 01f77564 (main) to their own hashes, and both pass check:catalog-drift at their heads, so each is internally consistent. Whichever merges first will define the resting order, and the second will rebase-regen through the artifact script. Non-blocking on this PR; loop-owner concern to sequence merges.

CI at head. Rollup SUCCESS. All Studio / Test / Producer / SDK / captures / catalog / lint / format / typecheck / build / windows / analyze / codeql green.

Body hygiene. Author's determinism claim + spotlight-clamp reasoning are both verified in the source. Independent adversarial review recorded (report on file). Rendered demo capture attached. [ ] Seen by home and Miguel before merge is your process, not a stamp gate.

Approving.

— tai