Add entry bridges (each naming a specific Level 1-3 artifact the
capability slots into) and literal exit bridges across the 7 Level 4
Substance pages, per the novice-to-advanced restructure plan.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Level 2 (anatomy, specification-dial, vocabulary, visual-specs, examples)
gets entry/exit bridges tying it into the Level 1 -> Level 3 arc, plus a
short note on the composition-structure mandate (0aaac7aa3 / PR #2599)
showing the framework enforces its own version of the prompt skeleton.
All existing prompts, renders, and tables are untouched.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Two real conflicts:
- docs/guides/prompting.mdx: modify/delete. This branch already deleted
it (383dd3568, replaced by the docs/prompting/* multi-page structure);
main separately fixed one line in the old file (cf7c1d760: lint+validate
-> the new consolidated `check` command). Kept the deletion — a
redirect to /prompting/overview already exists in docs.json — and
ported the same check-command fix forward into the two new-structure
pages that still had the stale lint+validate phrasing
(examples.mdx, rules-and-anti-patterns.mdx), so main's fix isn't lost.
- skills-manifest.json: auto-generated, regenerated fresh via
packages/cli/scripts/gen-skills-manifest.ts rather than hand-resolving
the JSON conflict markers.
Everything else auto-merged cleanly (no conflicts) — this pulls in
main's independent work since the branch diverged (media-use v2, the
check command consolidation, studio timeline updates, etc).
Co-Authored-By: Claude Opus <noreply@anthropic.com>
Learned from a strong example prompt (not reused verbatim — different
subject, same technique): a stop-motion/hand-animated look wants
irregular timing, but true unseeded randomness breaks HyperFrames'
determinism guarantee (same seek must always render the same frame).
The correct pattern is a PRNG seeded once at composition start, stepping
elements between held positions rather than tweening continuously —
distinct from rule 1's continuous ambient idle, which is a different
tool for a different texture.
Kept separate from "The grammar, measured" (rules 1-6, empirically
videoed) since there's no rendered proof for rule 7 yet — the new
section and example are illustrative, not claimed as verified.
Co-Authored-By: Claude Opus <noreply@anthropic.com>
* feat(cli): coordinate-frame layout findings in check
Four production compositions shipped with 100-600px layout drift, each a
different coordinate-frame confusion the check graded info or missed
entirely: viewport pixels written as container left/top, gsap x/y
treated as absolute position, a -350px margin fighting flex centering,
and stage-relative path coords drawn into a nested SVG.
Three new layout findings close the class:
- positioned_out_of_parent: an absolute/fixed element rendering mostly
outside its positioning ancestor (warning) — the parent needs no
overflow clipping, which is what let container_overflow miss it.
- box_out_of_canvas: a painted panel breaching the canvas (warning) —
text is canvas_overflow's, media is frame_out_of_frame's, painted
boxes were nobody's.
- connector_detached: a connector path whose endpoints land far from
every anchorable element (warning) — measured coordinates drawn into
an SVG with a different origin.
canvas_overflow additionally promotes from info to warning when held
across samples AND the breach exceeds 5% of the canvas.
All three are persistence-tiered and respect data-layout-allow-overflow.
Verified against the four incident compositions: every one now surfaces
its drift as held warnings (previously: info or silence).
* fix(cli): harden coordinate-frame findings against review false positives
Reworks all three findings after two-lens review (adversarial FP hunt in
real Chrome + maintainer pass):
- escaped_container (was positioned_out_of_parent): uses offsetParent
(transform-aware, skips fixed-as-canvas), exempts fully-detached
callouts within an attachment allowance while still flagging
touching-but-mostly-outside drift.
- panel_out_of_canvas (was box_out_of_canvas): paint alone qualifies
(flat solid panels were a false negative), fully off-canvas rects are
parked entrances and stay silent, pointer-events:none marks decorative
layers, hero-sized breaches warn while small bleeds stay info.
- connector_detached: endpoints via getPointAtLength + getScreenCTM
(viewBox, preserveAspectRatio, group transforms, every command type),
defs/marker/clipPath subtrees skipped, word-boundary connector naming,
containment tier limited to opaque non-ancestor targets (a text-bearing
wrapper contains its own diagram's endpoints).
- canvas_overflow promotion requires partial visibility — a fully
off-canvas rect is a parked entrance, not drift.
Verified: the four incident compositions still surface their drift as
held warnings; the review's false-positive repros (fixed HUD, callout,
parked entrance, corner bleed, marker arrowheads, g-transform and
viewBox-scaled connectors) are clean at warning level. Docs and the CLI
skill reference now describe the coordinate-frame findings.
* fix(cli): panel ownership is geometric — direct-text panels were a silent false negative
A painted panel whose direct text stays in-bounds while its box breaches
the canvas produced neither finding: canvas_overflow measures the text
range and panel_out_of_canvas skipped every own-text element. Skip the
panel finding only when the element's own text ALSO breaches (that
geometry belongs to canvas_overflow); pin the message/fixHint wording of
all three findings with positive assertions; document the SVG-internal
anchor blind spot.
* fix(cli): classify panel decoration by paint kind, not pointer-events
pointer-events:none exempted the framed-painting incident's gold frame
layers — hero content that happens to disable hit-testing. Decoration is
now gradient-only paint (spotlights, textures, vignettes); url() images,
solid fills and borders are content regardless of pointer-events.
* fix(cli): add fixHint to the test-local AuditIssue shape
* fix(cli): gradient stops decide content vs decoration; ownership matches canvas_overflow's tolerance
A gradient with any solid stop (alpha >= 0.6) is content — heroes and
cards painted with linear-gradient were invisible under the blanket
gradient exemption; all-translucent stops (spotlights, vignettes) stay
decoration. The text-ownership check now uses the audit tolerance that
canvas_overflow itself fires at, making the contract strict-mutex: any
text breach past that tolerance cedes the element, so a shallow 20px
text breach no longer double-reports.
Every other creation workflow gets either a dedicated page or an
explicit routing/contrast section inside a related one (website-to-
video inside product-launch.mdx, slideshow inside music-and-
slideshows.mdx). /remotion-to-hyperframes had neither — its only
mention anywhere in the 29-page guide was one line in overview.mdx's
install table.
Covers what to hand the agent (the exact source file, since a Remotion
project can register several Compositions), the one-way/no-reverse-
export scope, the lint-blocker categories that make the skill refuse
rather than approximate (useState/useEffect state machines, third-
party UI kits), and why SSIM validation against the Remotion original
is part of the deliverable, not optional. Added to the "By video type"
nav group in docs.json.
Co-Authored-By: Claude Opus <noreply@anthropic.com>
Scaffolded projects' npm run check now invokes the single check command
instead of chaining lint, validate, and inspect (three Chrome boots
become one). The CLI skill, its correctness reference, the entry skill's
capability map, README/docs catalog rows, the Mintlify CLI page (new
check section, deprecation banner on inspect), template CLAUDE/AGENTS
(byte-identical), root CLAUDE/AGENTS, and every creation-workflow skill
that taught the old sequence all point at check. snapshot keeps its
standalone sections; validate/inspect stay documented as deprecated
aliases with their check equivalents.
Validates 16 more published prompts end-to-end (variants across every
video type, caption styles, overlays with proven alpha, code diff and
highlight, edit verbs against the real attribute map, a generated-avatar
talking-head through embedded-captions, a presented slideshow deck) and
fixes every prompt flaw the builds surfaced. Adds keyword-to-render
grids for the easing, camera, and marker vocabulary; two more worked
visual specs with renders; a measured before/after pair proving the
motion grammar (frozen frames vs all-alive, 211KB vs 2.5MB); before/
after render pairs on editing and generated-artwork pages; and proof
renders on the overview.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Five-way review of the section against docs.json, catalog pages, and
skills. Fixes: stale /hyperframes-media skill row (real name /media-use)
plus missing music-to-video and slideshow workflow rows; caption
identity misattributions (keynote/documentary are scene-embedded, stomp
carries no rail); prompts violating the guide's own ambient-idle rule;
grounded keying rationale (u2net_human_seg); broken forward references;
nested-backtick code spans; brand casing and frontmatter dash
consistency throughout.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Moves all 28 example/validation renders to the docs' canonical media
host (s3://heygen-public via scripts/upload-docs-images.sh) and points
every embed at static.heygen.ai instead of HeyGenVerse asset URLs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Runs validation builds for the primary prompt on every new feature page
(all pass lint/validate/render) and applies the fixes they surfaced:
lower-third exit-length note, block-adaptation framing on the anatomy
technique slot, code-typing per-character wording, caption-component
demo-content note, map determinism/merge-at-source notes, device-mockup
demo-UI expectation, variables img clarification. Embeds 29 unedited
renders next to their prompts (HeyGenVerse-hosted), including a
default-vs-overridden pair demonstrating variables.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Adds 18 pages completing the Prompt Guide expansion (phases 2-4): six
video-type pages seeded by the verified example prompts, and twelve
feature pages giving prompting guidance for surfaces the docs implied
but never covered (transitions, caption styles, overlays, code blocks,
data/maps, VFX, rendering/output, editing existing videos, media/audio,
variables, runtimes/3D, design systems). All block/component/skill names
grounded against the catalog and skills; render pairing deferred.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Splits guides/prompting into a Prompt Guide nav group (overview + Fundamentals
+ Making it look good, 11 pages) with a redirect from the old URL. Applies
research-validated upgrades: beat content formula, before/after prompt rewrites
with engine rationale, single-variable iteration moves, absolute-target
calibration, camera/depth/pacing vocabulary, prerequisites gate.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Documentation-only updates surfaced by the prompt-guide validation
campaign. The behavior fixes from the same campaign are in a separate
PR (pipeline fixes); Tier 3 block bugs are tracked in #2107.
- hyperframes-core SKILL.md: new "First-pass lint gotchas" subsection —
the root composition element must carry data-start="0"
(root_composition_missing_data_start), and initial states belong in
gsap.fromTo, never a CSS transform paired with a GSAP tween on the
same property (gsap_css_transform_conflict). Both were guaranteed
first-build lint failures for agents writing compositions cold.
- media-use SKILL.md: "Weak local-model defaults" row corrected — local
Kokoro TTS is the free default with HeyGen as the paid upsell (was
described backwards), and bg-removal runs via the hyperframes CLI.
- docs/concepts/compositions.mdx: document that data-composition-src
paths resolve relative to the project root, not the referencing file.
- gsap-transforms-and-perf.md: runtime-created SVG (createElementNS)
has no reliable center pivot — draw geometry at local (0,0) inside a
positioning <g> so scale/rotation pivot on the shape's middle.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The old setup section was one long token-minting flow with the MCP
connector as an afterthought — but motion/shaders/storyboards need
only the connector (no token at all), and even brand tokens are
easier via MCP on any non-Enterprise plan. A reader wanting only
motion had to wade past REST scope tables meant for a different path.
Restructured into a decision table (what you want → which credential)
followed by two equal, independent steps. Step A's scope list now
states the 3 boxes to check on a normal (non-Enterprise) account
up front, instead of a generic 4-scope table the reader has to
interpret themselves. Step B now states the MCP connector's actual
capability (variable reads work on any plan, rate-limited by tier —
6 calls/month on Starter) instead of "no scopes to configure," which
undersold what it can do.
Co-Authored-By: Claude Opus <noreply@anthropic.com>
Miga's re-review caught it: the new RATE_LIMITED row (client auto-
retries with backoff) landed alongside the old pre-retry row ("wait a
minute and retry; chunk batch renders"), leaving two rows for the same
code — one accurate, one stale. Keep only the current one.
Co-Authored-By: Claude Opus <noreply@anthropic.com>
Addresses @miguel-heygen's review on #2112:
- BLOCKER: docs/guides/figma.mdx now matches the shipped code/skill —
adds the Library content: Read-only scope row (+ corrects the
'falls back, expected' line that was false without it), and the
troubleshooting table now says bad PATs surface as 403 Invalid token
(not 401), names the scope in FORBIDDEN, and documents RATE_LIMITED retry.
- nit: the batch summary line no longer claims '1 figma request' when every
node was a cache hit — says 'all reused from cache — no figma request'.
- nit: index.md regen moved to a finally, so a mid-batch RENDER_FAILED
leaves index.md consistent with the nodes that did freeze.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Vance Ingalls