stardust:prototype
For each directed page, render a proposed redesign as a
self-contained static HTML file at
stardust/prototypes/<slug>-proposed.html. Open the file in the
browser; iterate via chat-driven impeccable commands ("make the
hero bolder", "tighten the cup-note grid"). Mark approved once
the user signs off in the conversation.
prototype is not a renderer of its own design — it composes the
target spec written by direct (PRODUCT.md, DESIGN.md,
DESIGN.json, stardust/direction.md) onto the page content captured
by extract (stardust/current/pages/<slug>.json). Visual creativity
is delegated to $impeccable craft and the iteration commands
(bolder, quieter, distill, polish, colorize, typeset,
layout, adapt, animate, delight, overdrive, impeccable).
Inputs
<slug> — optional positional. Prototype just this page. Without
it, prototype every directed page that is not stale.
--all — prototype every directed page including stale ones.
--prep — optional. Run in migrate-prep mode: fill page-type
gaps (prototype one representative archetype per type) and, on
approval, write canon back to stardust/canon/ and
DESIGN.json.extensions.canon. See § Prep mode below and
reference/canon-extraction.md. Typically invoked via the
prepare-migration orchestrator.
--canon-from <slug> — optional. Override the default canon-
author (which is the first approved prototype, typically home).
Used when a different page should establish the design canon.
--publish-sample <slug> — submit the named slug to the
stardust showcase. Triggers the publish-sample sub-flow
documented in reference/publish-sample.md: eligibility checks,
file staging, PR creation against the upstream stardust repo.
Requires gh installed and authenticated. The showcase is a
visual demonstration, not a deployable site — placeholder
content is allowed and recorded in the PR body's § Unsourced
content section. Design-quality gates stay strict: refuses on
unjustified anti-toolbox hits, :root token contract failure,
data-attributes contract failure, or impeccable hard-rule
violations. P0/P1 critique findings warn but don't refuse.
The showcase publishes via GitHub Pages on merge.
--cinematic — optional. Layer a cinematic motion register
on top of the static prototype. The register is read from
DESIGN.json.extensions.motion.register (written by direct);
if absent, the prototype phase picks one using the same
heuristic per reference/motion-registers.md § Selection
heuristic. Output filename is <slug>-cinematic.html
(alongside the static <slug>-proposed.html, never replacing
it). Triggers the cinematic gates in motion validation
(reference/motion-validation.md § Pass 6).
--cinematic=<register> — optional. Same as --cinematic but
forces a specific register (arrival, kinetic-display,
live-systems, editorial, kinetic-grid). The override is
recorded in _provenance.motion.registerSource = "user-override"
so reviewers can spot when direction's heuristic was bypassed.
No opt-outs
prototype does not carry --no-* or --skip-* flags. The
quality gates (critique, audit, mobile-adapt audit, anti-toolbox
audit, content-sourcing scan) are the product — they're not
optional. If a gate refuses a file, the remediation is to fix
the file or override the gate by editing the file directly, not
to pass a flag that silently lowers the bar. Manual chat
overrides ("ship as-is", "accept the P1 findings") are still
available; the agent records the override verbatim in
_provenance so downstream consumers see the explicit
acknowledgement.
Setup
- Playwright re-probe (mandatory first step).
--no-save playwright
installs from earlier phases are pruned by any later real npm i
(extract SKILL.md § Setup → --no-save installs are ephemeral). Before
any rendering step, probe
node -e "import('playwright').then(()=>process.exit(0))" from the
project root and re-install (npm i -D playwright --no-save --legacy-peer-deps) on failure.
- Run the master skill's setup
(
skills/stardust/SKILL.md § Setup).
- Verify
stardust/state.json exists and contains at least one
directed page. If not, recommend $stardust direct and stop.
- Verify the project-root
DESIGN.md and DESIGN.json exist. If
not, the direction was not fully authored — recommend
$stardust direct and stop.
- Verify
stardust/direction.md has an active (not pending)
direction. Pending directions block prototype.
- Validate provenance on every page in scope. Call
validateProvenance(page) per
skills/stardust/reference/state-machine.md § Provenance
validation for every page that this run will render (the
single <slug> argument when present, otherwise every
non-stale directed/prototyped/approved page). Abort with
the helper's error when any page lacks live-render evidence
— re-running prototype against a synthesized page record
silently propagates the synthesis into the rendered prototype.
Surface Provenance OK on N pages once the check passes.
- Read
stardust/current/DESIGN.md (the descriptive snapshot of the
existing site, used as a fallback reference during render when the
proposed file needs to mirror an aspect of the captured surface).
Delegation mechanic
prototype does not author <slug>-proposed.html directly. The
heavy creative lift is delegated to $impeccable craft, and (when
needed) the structural plan to $impeccable shape. Spelling out the
mechanic matters because the carve-out documented in
skills/stardust/reference/artifact-map.md (where stardust authors
PRODUCT.md, DESIGN.md, DESIGN.json, current/PRODUCT.md,
current/DESIGN.md directly, treating impeccable's references as
format specs, not runtime commands) is load-bearing for those
five files only. It does NOT extend to:
stardust/prototypes/<slug>-proposed.html — must be authored by
$impeccable craft, not by stardust direct authoring.
- Iteration on the proposed file — must be driven through a
chat-driven invocation of an explicit impeccable command (per
the iteration paths section).
- Structural planning when a page is complex enough to need it —
$impeccable shape.
The proximate cause of past content fabrication was the agent
over-generalizing the direct-authoring carve-out to the proposed
HTML. Don't.
Invoking impeccable
When stardust runs in a Claude Code skill context (impeccable
exposed as the impeccable:impeccable Skill, not as a CLI), invoke
impeccable via the Skill tool with the sub-command and its args
mirroring the slash-command form:
Skill {
skill: "impeccable:impeccable",
args: "craft <feature-description>"
}
Sub-commands referenced from this skill are all routed through the
same Skill: craft, shape, plus the iteration commands
(bolder, quieter, distill, polish, colorize, typeset,
layout, adapt, animate, delight, overdrive, impeccable).
When impeccable is not available (CLI-only environments,
plugin uninstalled, sandbox without skill access), stop and tell
the user impeccable is required for prototype rendering. Recommend
installing the impeccable plugin. Do not fall back to direct
authoring of <slug>-proposed.html — the validation contract
craft enforces (anti-toolbox audit, divergence rules, type ratios,
content sourcing hierarchy) is not reproducible by direct
authoring, and falling back silently ships unverified output.
Stardust's job inside Phase 2 is therefore:
- Compose the inputs craft needs (page content from
current/pages/<slug>.json, target spec from DESIGN.md /
DESIGN.json, hard constraints from direction.md, content
sourcing rules from reference/proposed-file-shell.md § Content
sourcing hierarchy).
- Invoke craft via the Skill tool.
- Validate the result against the contract (
:root block, data
attributes, divergence audit, impeccable hard rules, content
sourcing). If validation fails, refuse to write — never paper over
craft output the agent thinks is "close enough."
The proposed file is whatever craft writes plus the validation
report; it is not stardust's authored artifact.
Procedure
Phase 1 — Plan the prototype (page-shape brief)
For each page in scope:
Read stardust/current/pages/<slug>.json for the page's structure
and content.
Read stardust/current/_brand-extraction.json for system
components and cross-promo data (the page's site-wide repeated
surfaces).
Read stardust/direction.md Active section for the resolved
direction, divergence inputs, and command sequence.
Read project-root DESIGN.md + DESIGN.json for the target
site system — tokens, abstract component vocabulary, named
system-component roles. The site system tells the agent what
the design language is; this Phase decides how it deploys to
this specific page.
Author stardust/prototypes/<slug>-shape.md — the per-page
compositional brief. Format spec:
reference/page-shape-brief.md. The brief carries the section
list, layout strategy, key states, interaction model, structural
data attributes, and the unsourced-content list (bridge to the
placeholder contract). Author directly — no interview, no
impeccable invocation; this is stardust's reasoning about how
the system deploys to this page given this content.
Show the brief to the user and wait for confirmation before
moving to Phase 2. The user can edit the brief in place
(rearrange sections, kill open questions, change composition
decisions); re-rendering Phase 2 will rebuild the proposed file
from the edited brief.
Hands-off compliance. When state.json.handsOff is true,
the shape brief is still authored and validated but the
user-confirmation wait is skipped — the brief is its own record.
Approval (Phase 5) under hands-off follows the mode's contract
defined in skills/stardust/SKILL.md § Hands-off mode.
$impeccable shape is not invoked in v0.2 (see
reference/page-shape-brief.md § Authoring procedure for the
rationale; revisit if per-page hand-authoring proves insufficient
across sites).
The brief decouples site-level concerns (in DESIGN.md) from
page-level deployment (per-page brief). A direction change
invalidates the system; existing briefs are content-aware-stale
only when the system change makes their composition impossible.
This recalibration of stale-flagging is documented in
skills/stardust/reference/state-machine.md § Stale flagging.
Brief-time disciplines (validator-enforced)
Five disciplines fire when the brief is authored. The brief
validator rejects briefs missing any of them; failure surfaces the
specific rule violated and stops Phase 1 before Phase 2 renders
anything. The disciplines exist to prevent the AI-slop failure mode
where a brief with only (DESIGN tokens) + (captured content) as
input produces template-shaped output regardless of brand.
Discipline 1 — Captured-source lineage per section. Every
section in ## Sections declares its captured-source origin. Forms:
- "subscription-card-on-bay (consolidates captured banner 2 +
dual-card right of
pages/home.json#landmarks[hero])" — derived
from one or more captured surfaces
- "site-wide system-component (carried from
_brand-extraction.json #systemComponents[kind=footer])" — chrome inherited from the
brand-surface
- "direction-authorized new" — composed against direction.md, not
derived from a captured source (require a one-line note naming
which direction movement justifies the new section)
Sections without lineage are rejected. The lineage list lives in
_provenance.capturedSourceLineage in the brief and propagates to
the rendered file.
Discipline 2 — Anti-template pass. For each captured component
pattern the page deploys (cards / banners / search rows / CTA bands
/ hero composition), the brief considers 2–3 layout alternatives
and picks the most differentiated. Curated default-patterns-to-escape
flags reflex picks:
- hero-then-bands silhouette (the universal AI silhouette)
- 5-up image-card grid as category nav
- nav-icon glyphs (search / cart / account) in a typographic
register
- centered-stack hero with two-button CTA pair
- captured-shape mirror-translated into new tokens
The brief justifies any reach for a default-pattern with a
captured-source citation that makes the pattern brand-appropriate
(e.g., "5-up grid preserved because the captured site's grid IS
the brand's signature catalogue shape; the alternative consideration
was a vertical ledger which the brand register rejects").
The alternatives SHOULD be reference-grounded when reference
research is available (per
skills/stardust/reference/reference-research.md): an alternative
cites a real reference screen in the entry's reference? field
using that file's evidence shape. Taste-only alternatives remain
valid when research is unavailable.
The audit lives in _provenance.antiTemplatePass[] with one entry
per captured pattern: { pattern, defaultReflex, alternatives[], picked, rationale, reference? }.
Discipline 3 — Surprise budget. The brief declares a surprise
field with one of: low | medium | high. Moves come from the bank
of non-template moves — see
skills/stardust/reference/divergence-toolkit.md § Non-template
move bank, with worked examples in
reference/anti-template-bank.md — or an evidence-shaped extension
per the bank's § Extension rule. Tier semantics:
low — brand-faithful + improvements only. Variant A's role
under reimagined; all of A1/A2/A3 under verbatim.
medium (default for B variants under reimagined) — one captured
cliché replaced by one move from the bank.
high (default for C variants under reimagined) — two clichés
replaced + one document-shape substitution.
Under ia-fidelity: verbatim (per
skills/stardust/reference/intent-dimensions.md § 9), the per-page
surprise budget is capped at low site-wide. The validator
refuses any verbatim-direction brief with surprise: medium or
high.
low ≠ generic. The budget bounds added divergence, not craft
or fidelity. low means brand-faithful + improvements + full
signature preservation, NOT "the most obvious faithful
interpretation." The recurring failure mode (moneyhub.com migration)
is the agent reading low / verbatim as license to strip the page
to a plain type-hero on a flat ground — the result is faithful but
forgettable and under-sells the redesign. Hold the craft bar at low:
keep the brand's distinctive elements, apply the improvements list,
and reproduce the signature.
Signature preservation is mandatory and budget-exempt. When the
captured page has a signature hero medium (background video / canvas /
WebGL / Lottie), signature motion (scroll / parallax / kinetic), or a
signature visual motif (per intent-dimensions.md § 8b), the brief
must reproduce it — with a static fallback, prefers-reduced- motion alternative, and (for overlaid text) a legibility scrim. This
does not consume the low allowance: carrying the brand's own
signature forward is fidelity, not divergence (§ 8b § Surprise-budget
exemption). Record the kept signatures in
_provenance.signatureElements[] as { kind, capturedSource, mechanism, fallback }. Render-refusal: a brief that flattens a
captured video/canvas/animation hero to a still, gradient, or
type-only hero — or drops a site-wide motif — is rejected at the
shape-brief audit; reproduce the signature instead.
Type-scale yield clause. When a tier-medium-or-higher variant's
captured-trait amplification structurally conflicts with a
brand-level type-scale rule from DESIGN.md (e.g. a
"Names-At-Headline-Scale" rule, or any other named type-scale
floor / ceiling), the brand-level rule may yield per-page. The
yield must be cited in _provenance.surpriseTier_typeScaleYields[]
with: { rule, variantDominantDimension, capturedTraitAmplified, yieldedTo, rationale }. The brand-level yield clause itself
(the named exception in DESIGN.md) is project-side, not
spec-side; the spec only requires the per-page citation when the
yield fires. See § Friction carve-out #4 below.
Discipline 4 — Substrate transitions are deliberate, named, and
rare. Default: single substrate across the page. Each substrate
exception requires a named purpose in the brief ("highlights the
featured-coffee promotional moment"). More than two substrate
transitions per page fails the brief. The transitions live in
_provenance.substrateTransitions as { default, exceptions[] }.
Exception — substrate-keyed document-shapes (friction carve-out
#2). When the variant's surprise: high move is a
substrate-keyed document-shape (zine, catalog-card,
poster-sequence), the per-section substrate IS the document-shape's
structural rhythm. The ≤ 2-transition cap does not apply; each
section's substrate must instead carry a per-section
captured-source citation (typically a per-SKU label color, a
per-page brand color, or a per-content-type ground convention from
the captured site). The exception is recorded in
_provenance.substrateTransitions.note with the document-shape
named and the citation source. The validator accepts > 2 transitions
when:
surprise: high is declared, AND
- the picked move from the bank is
document-shape with a
substrate-keyed sub-kind (zine / catalog-card / poster-sequence),
AND
- every transition in
exceptions[] carries a per-section
captured-source citation.
If any of the three conditions fails, the cap re-engages.
Discipline 5 — Heading hierarchy + voice classification tracked
per section. The brief requires:
- H1 declared once per page; subsequent sections H2; H3 children.
- Every literal value in copy classified as one of:
captured-verbatim, direction-authorized rewrite, or
placeholder. No section ships with copy without a classification.
The classification list lives in _provenance.voiceClassification[]
as { section, classification, copy?, source? } and propagates to
the rendered file.
Placeholder-ribbon labels (friction carve-out #3). System-
component honesty-signal labels (placeholder ribbons, "TBD" badges,
"from the brand team at migrate" markers) are direction-authorized
chrome, not placeholder content. They label placeholder prose /
data; the prose / data is what gets enumerated in
_provenance.unsourcedContent[]. The ribbon text itself does not.
Classify the ribbon labels as
direction-authorized chrome (per friction #3) so the placeholder
enumerator stays clean (no double-counting across every component
instance).
Phase 2 — Render the proposed page
Render stardust/prototypes/<slug>-proposed.html per
reference/proposed-file-shell.md § Required structure. Hard
requirements there:
:root token block as the first content of the first <style>
(per skills/stardust/reference/token-contract.md).
- Structural data attributes on every section (per
skills/stardust/reference/data-attributes.md).
- Provenance block as the first child of
<head>.
- Self-contained: no external CSS, no external JS.
- Content preserved from the current page (hero copy, CTAs, nav,
body) unless
direction.md authorises content changes.
- Content sourcing hierarchy (
reference/proposed-file-shell.md
§ Content sourcing hierarchy): every literal value rendered must
come from current/pages/<slug>.json, then voice samples, then
direction-authorised changes — or be rendered with the mandatory
PLACEHOLDER visual signature. Stats, addresses, quotes, tax IDs,
hours, prices, named-person words must never be invented. The
proposed file's _provenance.unsourcedContent[] lists every
placeholder so migrate can refuse to ship unverified content.
Delegate the heavy creative lift to $impeccable craft:
- Pass the page content and the resolved direction as the feature
description.
- Reference DESIGN.md / DESIGN.json as the design system.
- Pass
direction.md § Anti-references and § Divergence inputs as
hard constraints (so craft does not silently veer off the resolved
direction).
- Skip craft's "north star mock" generation step (direction.md is the
brief). Skip craft's "shape" call (already done if Phase 1 needed
it).
Modern-web-guidance consult. When the render implements
scroll-driven animation, view transitions, anchor positioning,
container queries, or perf-sensitive hero media, and the
modern-web-guidance plugin is installed, search it
(npx -y modern-web-guidance@latest search "<query>") and follow
the retrieved guide; cite the guide id in
_provenance.guidesConsulted[]. Skip silently when the plugin is
absent.
After craft returns, validate the output:
:root block present and complete (token-contract.md).
- Data attributes on every section (data-attributes.md).
- Anti-toolbox audit clean (each hit justified per divergence-toolkit.md
§ 1; record audit results in
DESIGN.json.extensions.divergence.anti_toolbox_hits
with the audit's amendments noted).
- Impeccable hard rules respected (OKLCH, type ratio ≥ 1.25, no
reflex slop).
- Content sourcing scan — every literal value in the rendered
output traces to one of the allowed sources
(
reference/proposed-file-shell.md § Content sourcing hierarchy).
Any value that doesn't is either wrapped in a [data-placeholder]
element with the mandatory visual signature, or the validation
fails. Build the _provenance.unsourcedContent[] list during
this scan.
If validation fails, do not write the file. Surface the failure to
the user with the specific rule violated and a suggested fix.
Craft-time disciplines (pre-write validators)
Four disciplines fire on the rendered file before it lands on
disk. These run after craft returns its output and before the file
is written; failure refuses the write with a substitute proposal
(Discipline 6) or a rule citation (7, 8), and Discipline 9 registers
detector ignores rather than refusing.
Discipline 6 — Reflex-reject font pre-flight. Grep the
declared font-family declarations against the reject list in
skills/stardust/reference/divergence-toolkit.md § Reflex-reject
fonts. If ≥ 3 reject-list families hit, refuse to write and
propose dimensionally-equivalent off-list substitutes from the
substitute table.
The check fires only on rendered files whose font-family
declarations the agent had freedom to pick. Mode A renders with
captured display + body families pinned (per
direct/SKILL.md § Mode A — Brand-faithful mode) bypass the
check — the inherited families are not a reflex choice. Record
the bypass reason in _provenance.reflexRejectAudit.bypassed with
the captured families named.
Discipline 7 — Variable-font axis engagement. When the resolved
direction's expressive or distinctiveness axes have moved past
their default (per
skills/stardust/reference/intent-dimensions.md §§ 2 + 5), engage
≥ 1 variable-font axis non-trivially per page using deck recipes
from the divergence toolkit. The toolkit's font decks expose named
axis recipes:
serif-luxury deck → Fraunces axis recipe: opsz 144, SOFT 100, WONK 1
bauhaus-functional deck → Bricolage axis recipe: opsz 96, wght 600
tactile-humanist deck → Recursive axis recipe: MONO 1, CRSV 0, CASL 0.5
Static-weight static-style across all type on a page that claims
expressive or distinctive movement fails the check.
Static-only-family precedence (friction carve-out #1). Family
substitution rules when the captured/pinned display font does not
ship a variable axis:
- If the captured/pinned display font is variable, engage its
axes per deck recipes (standard path).
- If the captured display font is static-only by family
(Bellfort, GT Sectra static, hand-lettered custom cuts whose
woff2 files don't expose a parseable fvar table without
runtime inspection), exempt the display family from the
check and engage the body family's axes instead (typically
wght and ital on a workhorse like Public Sans or Hanken
Grotesk).
- Document the family's static-only status in
DESIGN.json.extensions.divergence.font_deck.notes so downstream
pages don't re-run the check.
A pre-flight that blindly grep'd for font-variation-settings
would fire on every brand-faithful render of a static-only-display
brand. The exemption keeps the spirit of the discipline (engage
variable behavior where it exists) while honoring the letter when
the captured family literally has no axes to engage.
Off-deck accent fonts must ship with named expressive positions
(marginal-tag, pointer-scribble, headline-callout,
badge-fill — see toolkit § Font deck — expressive positions).
A face introduced without a position is a reflex pick, not an
expressive position.
Discipline 8 — Fidelity tier (--fidelity=quick|refined|production).
quick (default) — sketch fidelity. Brief decisions land;
micro-decisions stay provisional.
refined — adds a craft micro-pass per
reference/fidelity-refined-pass.md: formal type scale as CSS
custom properties, font-variant-numeric: tabular-nums on inline
digits, text-wrap: balance on headings + text-wrap: pretty
on body + hanging-punctuation: first on html, sliding
left-rule on nav hover, hairline hover on list items,
kicker + title + baseline-aligned more-link section-head triplet,
italic display couplet replacing structurally-vague <b>,
bottom-of-sidebar reworked as typographic plate.
production — refined + WCAG AA audit on every render + harden
pass (loading states, error states, content-overflow edges).
The tier is declared in the run invocation; persisted in
_provenance.fidelity. Default is quick.
Discipline 9 — Copy-cadence detector bypass under verbatim
fidelity. This extends the Mode-A reasoning of Discipline 6 from
fonts to prose. impeccable's design detector ships prose-voice rules
(em-dash-overuse, marketing-buzzword, and similar copy-cadence
checks) that assume the copy is the agent's to rewrite. Under
ia-fidelity: verbatim — or any faithful/Mode-A render where the body
copy is captured-verbatim — that assumption is false by
construction: the prose is the source brand's, reproduced exactly per
the content-sourcing hierarchy, and rewriting it to satisfy a cadence
rule is the fabrication the fidelity setting exists to prevent. So
when the rendered file's copy classification is captured-verbatim
(per Discipline 5's voiceClassification), register those copy-cadence
rules as intentional ignores for the <slug>-proposed.html files
before the design hook fires — the same way Discipline 6 bypasses the
font reflex-reject check for pinned families. Scope the ignore to the
proposed files only, never to the project's own source (blocks,
styles, components), where the rules still apply because that copy
is the agent's. Record the bypass in
_provenance.copyCadenceBypass with the rules ignored and the
classification basis. The 2026-06-26 knack.com run hit this: the hook
flagged em-dashes and "enterprise-grade" on Knack's own headings
("Built on Enterprise-Grade Components") under a verbatim direction,
and the only correct response was to leave the captured copy untouched
and record the bypass.
Bound the bypass: it covers prose-cadence rules only. Structural and
craft detector rules (design-system-radius, contrast failures,
reflex layout slop) are not exempted by verbatim fidelity — those
govern the agent's own CSS and structure, which faithful mode does not
freeze. Listing a rule under this bypass requires it be a copy-voice
rule whose subject is the captured prose.
Phase 2.4 — Motion application (when --cinematic)
Fires only when --cinematic (with or without an explicit
register) was passed, OR when
DESIGN.json.extensions.motion.register was authored by direct
and the user did not opt out. Produces
stardust/prototypes/<slug>-cinematic.html alongside the static
<slug>-proposed.html — the static prototype is never replaced.
Procedure:
Resolve the register. For single-variant runs, read
DESIGN.json.extensions.motion.register. For multi-variant
runs (when DESIGN-<id>.json files exist at the project
root), read DESIGN-<id>.json.extensions.motion.register for
the variant currently being rendered — each variant may carry
its own register, or omit it entirely (variant renders
static). If --cinematic=<register> was passed at the CLI,
the CLI value wins only for the variants in scope (every
variant when no <slug> filter is set, otherwise the
variant(s) matching the filter).
Record the source in _provenance.motion.registerSource
("direct" for per-variant DESIGN-authored, "user-override"
for CLI). If neither path resolved a register for this
variant, fall through to the selection heuristic in
reference/motion-registers.md § Selection heuristic. If the
heuristic itself returns no register (e.g. the variant's
PRODUCT.md Brand Personality maps to none of the five
registers, and no evidence-shaped extension register applies
per the bank's § Extension rule in
reference/motion-registers.md), skip Phase 2.4 entirely for
that variant — render it static.
The per-variant resolution is what lets uplift produce a
three-variant set where only variant C engages motion: direct
writes the register into DESIGN-C.json only, and Phase 2.4
fires for C alone (A and B render static).
Stage Lenis assets. Copy
skills/prototype/assets/motion/lenis.min.js and lenis.min.css
into stardust/prototypes/ (idempotent — skip if shas match).
Cinematic prototypes load these via relative paths.
Read the canonical runtime. Embed the inline script from
reference/motion-runtime.md § The canonical script verbatim,
with the animConfig constants rewritten per the active
register's token defaults (reference/motion-registers.md §
The five registers § Token defaults).
Layer the register's CSS. Append the register-specific
keyframes and class rules (entrance keyframes, parallax CSS
custom properties, marquee animations, pulse animations) to
the file's <style> block. The set of keyframes is closed
per register; see reference/motion-runtime.md § Per-register
tuning.
Annotate target HTML. Walk the rendered DOM and emit the
motion data-* attributes per the register's vocabulary:
arrival: [data-anim] on section heads, body copy,
CTAs, and tile cards; [data-countup] on numeric values;
[data-parallax] on the hero photograph.
kinetic-display: [data-anim] on most sections; [data-split]
on display-cap headlines (<h1 data-split>DINE</h1>);
[data-flip] on terminal codes / gate numbers; .word
spans on display headlines for clip-path word wipes.
live-systems: [data-tile-anim] on every ops-tile and
card; [data-countup] on every numeric data value;
[data-fill] on every bar inner element; .live-sweep
on the live-data container; .marquee__track on the top
ticker.
editorial: [data-anim] only; [data-parallax] with
reduced magnitude on hero imagery; never [data-flip] /
[data-fill] / [data-split].
kinetic-grid: [data-tile-anim] on cards; [data-anim]
on section heads.
Full per-register attribute matrix: reference/motion-registers.md
§ Data-attributes consumed.
Inject the <noscript> fallback. Add the no-JS override
from reference/motion-runtime.md § No-JS fallback to <head>
so the file degrades to its static-end state without
JavaScript.
Update _provenance. Add the motion block:
"motion": {
"register": "<register-name>",
"registerSource": "direct | user-override | heuristic | extension",
"runtimeVersion": "v1",
"lenisAssets": { "js": "lenis.min.js", "css": "lenis.min.css" },
"attributesEmitted": ["data-anim", "data-countup", ...]
}
Hand off to Phase 2.8 (motion validation). The cinematic
gates in reference/motion-validation.md § Pass 6 fire
automatically because the rendered file declares the
_provenance.motion.register field.
The static <slug>-proposed.html is unaffected by this phase.
Both files are reviewable; the brand owner sees the cinematic
version when motion is part of the redesign brief, the static
version when migration / accessibility audit is the focus.
Output paths
| File |
Owner phase |
When |
stardust/prototypes/<slug>-proposed.html |
Phase 2 (always) |
Always written. |
stardust/prototypes/<slug>-cinematic.html |
Phase 2.4 (--cinematic) |
Written alongside; static remains. |
stardust/prototypes/lenis.min.js |
Phase 2.4 (--cinematic) |
Copied from skill assets. |
stardust/prototypes/lenis.min.css |
Phase 2.4 (--cinematic) |
Copied from skill assets. |
When to use the static-only path
The static prototype remains the load-bearing artifact for:
- Brand-faithful inheritance reviews (motion is additive — the
static prototype is the canonical "yes, that's us, refreshed"
surface).
- Accessibility audits (motion-driven pages are harder to evaluate
in their reduced-motion state).
- Migration consumption (
migrate reads the static prototype as
its primary source. It does not merge the cinematic layer —
per skills/migrate/SKILL.md § Phase 2 → Cinematic sibling, it
carries the motion assets (lenis.min.*) through to
migrated/assets/motion/ and records
cinematic-variant-not-consumed in the sidecar).
The static prototype must pass every gate independently — the
cinematic layer cannot rescue a static prototype that fails
Phases 2.5–2.7.
Phases 2.5 – 2.8 — Quality gates: Critique → Audit → Adapt → Motion (Discipline 9)
Four mandatory gate phases run by default before any prototype
can advance to prototyped. They implement Discipline 9: critique
covers design, audit covers technical correctness, adapt
covers viewport behaviour, motion covers scroll-driven and
time-driven choreography correctness. P0/P1 findings from any
of the four block prototyped until acknowledged. None have an
opt-out flag (per § No opt-outs).
| Sub-phase |
Focus |
Catches |
| 2.5 Critique |
Design judgment |
AI-slop reflexes, hierarchy regressions, contrast / cognitive issues, register drift |
| 2.6 Audit |
Technical correctness |
a11y (alt, focus, contrast ratios computed), responsive overflow at 4–6 viewports, performance (LCP, image weights), JS-dependent-hidden-state |
| 2.7 Adapt |
Viewport behaviour |
doc-width · overflow · sticky · grid columns · font scaling at 1920 / 1440 / 1280 / 800 / 414 / 375, mobile-nav-collapse audit |
| 2.8 Motion |
Scroll / time-driven correctness |
clipped-container reveal timing, animation-range vs reading position, anim-enter trigger reachability, reduced-motion override completeness, no-JS fallback, multi-viewport scroll-driven check |
Phase 2.8 fires only when the rendered file declares ≥ 1 named
choreography (per the page-shape brief's motion stack) OR uses
animation-timeline:, @scroll-timeline, IntersectionObserver-driven
entry triggers, or rAF loops reading getBoundingClientRect(),
OR when Phase 2.4 (motion application under --cinematic)
emitted a <slug>-cinematic.html file with motion attributes per
reference/motion-attributes.md. Cinematic prototypes additionally
trigger the cinematic-mode gates in reference/motion-validation.md
§ Pass 6 (Lenis bootstrap, reduced-motion fallback completeness,
scroll-jack check, three-position screenshots, register-match
audit, motion C-cliff detector). Static prototypes skip the
discipline entirely.
The wasatch dry-run on 2026-05-13 caught two real bugs the brief
and craft phases missed (a WCAG miscalculation off by 0.27–1.86
points on multiple pairs of /beers variant C, and an LCP image
lazy-loaded on the first catalog entry). Both were caught by the
audit half — not by critique. The detector's contrast
computation and responsive performance check are the load-bearing
audits; without them the file would have shipped with quantifiable
WCAG failures.
Phase 2.5 — Critique
Before opening the proposed file in the browser, run two parallel
validators against the rendered proposed file: critique and audit. They
are explicitly designed as a complementary pair — critique
covers design (AI-slop reflexes, hierarchy, brand fit,
cognitive load); audit covers technical correctness
(accessibility / performance / theming / responsive /
anti-patterns). Running only critique misses every quantifiable
WCAG / perf / responsive failure; running only audit misses
brand-misalignment and design slop. The pass is a contract,
not a courtesy.
The 2026-05-04 nvidia.com home prototype critique returned
1 P0 + 2 P1 + 3 P2; the audit on the same artifact returned
six additional findings (no skip-link, theme carousels
without keyboard arrow nav, hero ~3.5MB without responsive
<picture>, layout-property animation, JS-gated reveal with
no <noscript> fallback, scroll-behavior: smooth not
respecting prefers-reduced-motion). None were design issues,
none would have been caught by critique alone. Without an
audit gate the page would have been marked prototyped with
quantifiable WCAG failures.
Procedure:
Run both validators in parallel. Invoke impeccable:impeccable
twice in the same Skill-tool batch:
Skill { skill: "impeccable:impeccable",
args: "critique stardust/prototypes/<slug>-proposed.html --json" }
Skill { skill: "impeccable:impeccable",
args: "audit stardust/prototypes/<slug>-proposed.html --json" }
Each returns a JSON findings list — each finding has
priority (P0 / P1 / P2 / P3), category (hierarchy /
contrast / motion / a11y / perf / responsive / etc.), and a
one-line description. Capture critique findings into
_provenance.critique[] and audit findings into
_provenance.audit[] on the proposed file (append; never
overwrite previous runs' entries).
Brand-faithful inversion auto-dismiss. Both validators
ship known false positives on Mode A renders — Arial fallback
reads as "overused-font," eyebrow uppercase reads as
"all-caps body," pure white / pure black flagged when the
brand's captured palette includes them. Before surfacing
findings to the user, diff each finding against
DESIGN.json#extensions.divergence.brand_faithful_inversions[]
and DESIGN.md#narrative.rules (e.g. permitted uppercase
contexts). Drop findings whose category and target match an
approved inversion; keep the original list in
_provenance.<critique|audit>[] with a
dismissedAsBrandFaithful: true flag for audit-trail
purposes. The user-facing report shows only the real hits.
Vision gate. Render a screenshot of the proposed file and
study it NEXT TO the captured source screenshot
(stardust/current/assets/screenshots/<slug>.png when present).
Judge visually, not from the DOM: brand-fit (would the
brand owner say "that's us"?), signature preservation (hero
medium / motif carried, per Discipline 3's signature clause),
and hierarchy at a glance (does the eye land where the
brief says it should?). Record
_provenance.visionCheck = { verdict: pass|fail, observations[] };
a fail is a P1 finding through the same gat
…(truncated)
1---2name: prototype3description: Render a proposed redesign of a page on the current website as a self-contained static HTML file, then iterate via the impeccable craft loop. Per-page, idempotent, stale-aware. Use when the user asks for a redesign prototype, a before/after comparison, a design preview, a page mockup, a visual diff of the redesign, or invokes /stardust:prototype.4license: Apache-2.05metadata:6 source: https://skillmd.com/skills/adobe/prototype7---89# stardust:prototype1011For each `directed` page, render a **proposed redesign** as a12self-contained static HTML file at13`stardust/prototypes/<slug>-proposed.html`. Open the file in the14browser; iterate via chat-driven impeccable commands ("make the15hero bolder", "tighten the cup-note grid"). Mark `approved` once16the user signs off in the conversation.1718`prototype` is not a renderer of its own design — it composes the19target spec written by `direct` (`PRODUCT.md`, `DESIGN.md`,20`DESIGN.json`, `stardust/direction.md`) onto the page content captured21by `extract` (`stardust/current/pages/<slug>.json`). Visual creativity22is delegated to `$impeccable craft` and the iteration commands23(`bolder`, `quieter`, `distill`, `polish`, `colorize`, `typeset`,24`layout`, `adapt`, `animate`, `delight`, `overdrive`, `impeccable`).2526## Inputs2728- `<slug>` — optional positional. Prototype just this page. Without29 it, prototype every `directed` page that is not `stale`.30- `--all` — prototype every `directed` page including stale ones.31- `--prep` — optional. Run in **migrate-prep mode**: fill page-type32 gaps (prototype one representative archetype per type) and, on33 approval, write canon back to `stardust/canon/` and34 `DESIGN.json.extensions.canon`. See § Prep mode below and35 `reference/canon-extraction.md`. Typically invoked via the36 `prepare-migration` orchestrator.37- `--canon-from <slug>` — optional. Override the default canon-38 author (which is the first approved prototype, typically `home`).39 Used when a different page should establish the design canon.40- `--publish-sample <slug>` — submit the named slug to the41 stardust showcase. Triggers the publish-sample sub-flow42 documented in `reference/publish-sample.md`: eligibility checks,43 file staging, PR creation against the upstream stardust repo.44 Requires `gh` installed and authenticated. The showcase is a45 visual demonstration, not a deployable site — placeholder46 content is allowed and recorded in the PR body's § Unsourced47 content section. Design-quality gates stay strict: refuses on48 unjustified anti-toolbox hits, `:root` token contract failure,49 data-attributes contract failure, or impeccable hard-rule50 violations. P0/P1 critique findings warn but don't refuse.51 The showcase publishes via GitHub Pages on merge.52- `--cinematic` — optional. Layer a cinematic motion register53 on top of the static prototype. The register is read from54 `DESIGN.json.extensions.motion.register` (written by `direct`);55 if absent, the prototype phase picks one using the same56 heuristic per `reference/motion-registers.md` § Selection57 heuristic. Output filename is `<slug>-cinematic.html`58 (alongside the static `<slug>-proposed.html`, never replacing59 it). Triggers the cinematic gates in motion validation60 (`reference/motion-validation.md` § Pass 6).61- `--cinematic=<register>` — optional. Same as `--cinematic` but62 forces a specific register (`arrival`, `kinetic-display`,63 `live-systems`, `editorial`, `kinetic-grid`). The override is64 recorded in `_provenance.motion.registerSource = "user-override"`65 so reviewers can spot when direction's heuristic was bypassed.6667### No opt-outs6869`prototype` does not carry `--no-*` or `--skip-*` flags. The70quality gates (critique, audit, mobile-adapt audit, anti-toolbox71audit, content-sourcing scan) are the product — they're not72optional. If a gate refuses a file, the remediation is to fix73the file or override the gate by editing the file directly, not74to pass a flag that silently lowers the bar. Manual chat75overrides ("ship as-is", "accept the P1 findings") are still76available; the agent records the override verbatim in77`_provenance` so downstream consumers see the explicit78acknowledgement.7980## Setup81820. **Playwright re-probe (mandatory first step).** `--no-save` playwright83 installs from earlier phases are pruned by any later real `npm i`84 (extract SKILL.md § Setup → `--no-save` installs are ephemeral). Before85 any rendering step, probe86 `node -e "import('playwright').then(()=>process.exit(0))"` from the87 project root and re-install (`npm i -D playwright --no-save88 --legacy-peer-deps`) on failure.891. Run the master skill's setup90 (`skills/stardust/SKILL.md` § Setup).912. Verify `stardust/state.json` exists and contains at least one92 `directed` page. If not, recommend `$stardust direct` and stop.933. Verify the project-root `DESIGN.md` and `DESIGN.json` exist. If94 not, the direction was not fully authored — recommend95 `$stardust direct` and stop.964. Verify `stardust/direction.md` has an active (not pending)97 direction. Pending directions block prototype.985. **Validate provenance on every page in scope.** Call99 `validateProvenance(page)` per100 `skills/stardust/reference/state-machine.md` § Provenance101 validation for every page that this run will render (the102 single `<slug>` argument when present, otherwise every103 non-stale `directed`/`prototyped`/`approved` page). Abort with104 the helper's error when any page lacks live-render evidence105 — re-running `prototype` against a synthesized page record106 silently propagates the synthesis into the rendered prototype.107 Surface `Provenance OK on N pages` once the check passes.1086. Read `stardust/current/DESIGN.md` (the descriptive snapshot of the109 existing site, used as a fallback reference during render when the110 proposed file needs to mirror an aspect of the captured surface).111112## Delegation mechanic113114`prototype` does **not** author `<slug>-proposed.html` directly. The115heavy creative lift is delegated to `$impeccable craft`, and (when116needed) the structural plan to `$impeccable shape`. Spelling out the117mechanic matters because the carve-out documented in118`skills/stardust/reference/artifact-map.md` (where stardust authors119`PRODUCT.md`, `DESIGN.md`, `DESIGN.json`, `current/PRODUCT.md`,120`current/DESIGN.md` directly, treating impeccable's references as121*format specs*, not runtime commands) is **load-bearing for those122five files only**. It does NOT extend to:123124- `stardust/prototypes/<slug>-proposed.html` — must be authored by125 `$impeccable craft`, not by stardust direct authoring.126- Iteration on the proposed file — must be driven through a127 chat-driven invocation of an explicit impeccable command (per128 the iteration paths section).129- Structural planning when a page is complex enough to need it —130 `$impeccable shape`.131132The proximate cause of past content fabrication was the agent133over-generalizing the direct-authoring carve-out to the proposed134HTML. Don't.135136### Invoking impeccable137138When stardust runs in a Claude Code skill context (impeccable139exposed as the `impeccable:impeccable` Skill, not as a CLI), invoke140impeccable via the Skill tool with the sub-command and its args141mirroring the slash-command form:142143```144Skill {145 skill: "impeccable:impeccable",146 args: "craft <feature-description>"147}148```149150Sub-commands referenced from this skill are all routed through the151same Skill: `craft`, `shape`, plus the iteration commands152(`bolder`, `quieter`, `distill`, `polish`, `colorize`, `typeset`,153`layout`, `adapt`, `animate`, `delight`, `overdrive`, `impeccable`).154155When impeccable is **not** available (CLI-only environments,156plugin uninstalled, sandbox without skill access), stop and tell157the user impeccable is required for prototype rendering. Recommend158installing the impeccable plugin. Do not fall back to direct159authoring of `<slug>-proposed.html` — the validation contract160craft enforces (anti-toolbox audit, divergence rules, type ratios,161content sourcing hierarchy) is not reproducible by direct162authoring, and falling back silently ships unverified output.163164Stardust's job inside Phase 2 is therefore:165166- Compose the inputs craft needs (page content from167 `current/pages/<slug>.json`, target spec from `DESIGN.md` /168 `DESIGN.json`, hard constraints from `direction.md`, content169 sourcing rules from `reference/proposed-file-shell.md` § Content170 sourcing hierarchy).171- Invoke craft via the Skill tool.172- Validate the result against the contract (`:root` block, data173 attributes, divergence audit, impeccable hard rules, content174 sourcing). If validation fails, refuse to write — never paper over175 craft output the agent thinks is "close enough."176177The proposed file is whatever craft writes plus the validation178report; it is not stardust's authored artifact.179180## Procedure181182### Phase 1 — Plan the prototype (page-shape brief)183184For each page in scope:1851861. Read `stardust/current/pages/<slug>.json` for the page's structure187 and content.1882. Read `stardust/current/_brand-extraction.json` for system189 components and cross-promo data (the page's site-wide repeated190 surfaces).1913. Read `stardust/direction.md` Active section for the resolved192 direction, divergence inputs, and command sequence.1934. Read project-root `DESIGN.md` + `DESIGN.json` for the target194 site system — tokens, abstract component vocabulary, named195 system-component roles. The site system tells the agent *what196 the design language is*; this Phase decides *how it deploys to197 this specific page*.1985. **Author `stardust/prototypes/<slug>-shape.md`** — the per-page199 compositional brief. Format spec:200 `reference/page-shape-brief.md`. The brief carries the section201 list, layout strategy, key states, interaction model, structural202 data attributes, and the unsourced-content list (bridge to the203placeholder contract). Author directly — no interview, no204 impeccable invocation; this is stardust's reasoning about how205 the system deploys to this page given this content.2066. Show the brief to the user and wait for confirmation before207 moving to Phase 2. The user can edit the brief in place208 (rearrange sections, kill open questions, change composition209 decisions); re-rendering Phase 2 will rebuild the proposed file210 from the edited brief.211212 **Hands-off compliance.** When `state.json.handsOff` is true,213 the shape brief is still authored and validated but the214 user-confirmation wait is skipped — the brief is its own record.215 Approval (Phase 5) under hands-off follows the mode's contract216 defined in `skills/stardust/SKILL.md` § Hands-off mode.217218`$impeccable shape` is **not** invoked in v0.2 (see219`reference/page-shape-brief.md` § Authoring procedure for the220rationale; revisit if per-page hand-authoring proves insufficient221across sites).222223The brief decouples site-level concerns (in DESIGN.md) from224page-level deployment (per-page brief). A direction change225invalidates the system; existing briefs are content-aware-stale226only when the system change makes their composition impossible.227This recalibration of stale-flagging is documented in228`skills/stardust/reference/state-machine.md` § Stale flagging.229230#### Brief-time disciplines (validator-enforced)231232Five disciplines fire when the brief is authored. The brief233validator rejects briefs missing any of them; failure surfaces the234specific rule violated and stops Phase 1 before Phase 2 renders235anything. The disciplines exist to prevent the AI-slop failure mode236where a brief with only `(DESIGN tokens) + (captured content)` as237input produces template-shaped output regardless of brand.238239**Discipline 1 — Captured-source lineage per section.** Every240section in `## Sections` declares its captured-source origin. Forms:241242- *"subscription-card-on-bay (consolidates captured banner 2 +243 dual-card right of `pages/home.json#landmarks[hero]`)"* — derived244 from one or more captured surfaces245- *"site-wide system-component (carried from `_brand-extraction.json246 #systemComponents[kind=footer]`)"* — chrome inherited from the247 brand-surface248- *"direction-authorized new"* — composed against direction.md, not249 derived from a captured source (require a one-line note naming250 which direction movement justifies the new section)251252Sections without lineage are rejected. The lineage list lives in253`_provenance.capturedSourceLineage` in the brief and propagates to254the rendered file.255256**Discipline 2 — Anti-template pass.** For each captured component257pattern the page deploys (cards / banners / search rows / CTA bands258/ hero composition), the brief considers 2–3 layout alternatives259and picks the most differentiated. Curated default-patterns-to-escape260flags reflex picks:261262- hero-then-bands silhouette (the universal AI silhouette)263- 5-up image-card grid as category nav264- nav-icon glyphs (search / cart / account) in a typographic265 register266- centered-stack hero with two-button CTA pair267- captured-shape mirror-translated into new tokens268269The brief justifies any reach for a default-pattern with a270captured-source citation that makes the pattern brand-appropriate271(e.g., *"5-up grid preserved because the captured site's grid IS272the brand's signature catalogue shape; the alternative consideration273was a vertical ledger which the brand register rejects"*).274275The alternatives SHOULD be reference-grounded when reference276research is available (per277`skills/stardust/reference/reference-research.md`): an alternative278cites a real reference screen in the entry's `reference?` field279using that file's evidence shape. Taste-only alternatives remain280valid when research is unavailable.281282The audit lives in `_provenance.antiTemplatePass[]` with one entry283per captured pattern: `{ pattern, defaultReflex, alternatives[],284picked, rationale, reference? }`.285286**Discipline 3 — Surprise budget.** The brief declares a `surprise`287field with one of: `low | medium | high`. Moves come from the bank288of non-template moves — see289`skills/stardust/reference/divergence-toolkit.md` § Non-template290move bank, with worked examples in291`reference/anti-template-bank.md` — or an evidence-shaped extension292per the bank's § Extension rule. Tier semantics:293294- `low` — brand-faithful + improvements only. Variant A's role295 under reimagined; all of A1/A2/A3 under verbatim.296- `medium` (default for B variants under reimagined) — one captured297 cliché replaced by one move from the bank.298- `high` (default for C variants under reimagined) — two clichés299 replaced + one document-shape substitution.300301Under `ia-fidelity: verbatim` (per302`skills/stardust/reference/intent-dimensions.md` § 9), the per-page303surprise budget is **capped at `low` site-wide**. The validator304refuses any verbatim-direction brief with `surprise: medium` or305`high`.306307**`low` ≠ generic.** The budget bounds *added* divergence, not craft308or fidelity. `low` means **brand-faithful + improvements + full309signature preservation**, NOT "the most obvious faithful310interpretation." The recurring failure mode (moneyhub.com migration)311is the agent reading `low` / `verbatim` as license to strip the page312to a plain type-hero on a flat ground — the result is faithful but313forgettable and under-sells the redesign. Hold the craft bar at `low`:314keep the brand's distinctive elements, apply the improvements list,315and reproduce the signature.316317**Signature preservation is mandatory and budget-exempt.** When the318captured page has a signature hero medium (background video / canvas /319WebGL / Lottie), signature motion (scroll / parallax / kinetic), or a320signature visual motif (per `intent-dimensions.md` § 8b), the brief321**must** reproduce it — with a static fallback, `prefers-reduced-322motion` alternative, and (for overlaid text) a legibility scrim. This323does **not** consume the `low` allowance: carrying the brand's own324signature forward is fidelity, not divergence (§ 8b § Surprise-budget325exemption). Record the kept signatures in326`_provenance.signatureElements[]` as `{ kind, capturedSource,327mechanism, fallback }`. **Render-refusal:** a brief that flattens a328captured video/canvas/animation hero to a still, gradient, or329type-only hero — or drops a site-wide motif — is rejected at the330shape-brief audit; reproduce the signature instead.331332**Type-scale yield clause.** When a tier-`medium`-or-higher variant's333captured-trait amplification structurally conflicts with a334brand-level type-scale rule from `DESIGN.md` (e.g. a335"Names-At-Headline-Scale" rule, or any other named type-scale336floor / ceiling), the brand-level rule may yield per-page. The337yield must be cited in `_provenance.surpriseTier_typeScaleYields[]`338with: `{ rule, variantDominantDimension, capturedTraitAmplified,339yieldedTo, rationale }`. The brand-level yield clause itself340(the named exception in `DESIGN.md`) is project-side, not341spec-side; the spec only requires the per-page citation when the342yield fires. See § Friction carve-out #4 below.343344**Discipline 4 — Substrate transitions are deliberate, named, and345rare.** Default: single substrate across the page. Each substrate346exception requires a named purpose in the brief (*"highlights the347featured-coffee promotional moment"*). **More than two substrate348transitions per page fails the brief.** The transitions live in349`_provenance.substrateTransitions` as `{ default, exceptions[] }`.350351**Exception — substrate-keyed document-shapes (friction carve-out352#2).** When the variant's `surprise: high` move is a353substrate-keyed document-shape (zine, catalog-card,354poster-sequence), the per-section substrate IS the document-shape's355structural rhythm. The ≤ 2-transition cap does not apply; each356section's substrate must instead carry a **per-section357captured-source citation** (typically a per-SKU label color, a358per-page brand color, or a per-content-type ground convention from359the captured site). The exception is recorded in360`_provenance.substrateTransitions.note` with the document-shape361named and the citation source. The validator accepts > 2 transitions362when:3633641. `surprise: high` is declared, AND3652. the picked move from the bank is `document-shape` with a366 substrate-keyed sub-kind (zine / catalog-card / poster-sequence),367 AND3683. every transition in `exceptions[]` carries a per-section369 captured-source citation.370371If any of the three conditions fails, the cap re-engages.372373**Discipline 5 — Heading hierarchy + voice classification tracked374per section.** The brief requires:375376- H1 declared once per page; subsequent sections H2; H3 children.377- Every literal value in copy classified as one of:378 `captured-verbatim`, `direction-authorized rewrite`, or379 `placeholder`. No section ships with copy without a classification.380381The classification list lives in `_provenance.voiceClassification[]`382as `{ section, classification, copy?, source? }` and propagates to383the rendered file.384385**Placeholder-ribbon labels (friction carve-out #3).** System-386component honesty-signal labels (placeholder ribbons, "TBD" badges,387"from the brand team at migrate" markers) are **direction-authorized388chrome**, not placeholder content. They label placeholder prose /389data; the prose / data is what gets enumerated in390`_provenance.unsourcedContent[]`. The ribbon text itself does not.391Classify the ribbon labels as392`direction-authorized chrome (per friction #3)` so the placeholder393enumerator stays clean (no double-counting across every component394instance).395396### Phase 2 — Render the proposed page397398Render `stardust/prototypes/<slug>-proposed.html` per399`reference/proposed-file-shell.md` § Required structure. Hard400requirements there:401402- `:root` token block as the first content of the first `<style>`403 (per `skills/stardust/reference/token-contract.md`).404- Structural data attributes on every section (per405 `skills/stardust/reference/data-attributes.md`).406- Provenance block as the first child of `<head>`.407- Self-contained: no external CSS, no external JS.408- The site's favicon inlined in the head as a data: URI when409 `stardust/current/assets/favicon.<ext>` exists (per410 `reference/proposed-file-shell.md` § Hard requirements #8).411- Content preserved from the current page (hero copy, CTAs, nav,412 body) unless `direction.md` authorises content changes.413- **Content sourcing hierarchy** (`reference/proposed-file-shell.md`414 § Content sourcing hierarchy): every literal value rendered must415 come from `current/pages/<slug>.json`, then voice samples, then416 direction-authorised changes — or be rendered with the mandatory417 PLACEHOLDER visual signature. Stats, addresses, quotes, tax IDs,418 hours, prices, named-person words must never be invented. The419 proposed file's `_provenance.unsourcedContent[]` lists every420 placeholder so migrate can refuse to ship unverified content.421422Delegate the heavy creative lift to `$impeccable craft`:423424- Pass the page content and the resolved direction as the feature425 description.426- Reference DESIGN.md / DESIGN.json as the design system.427- Pass `direction.md` § Anti-references and § Divergence inputs as428 hard constraints (so craft does not silently veer off the resolved429 direction).430- Skip craft's "north star mock" generation step (direction.md is the431 brief). Skip craft's "shape" call (already done if Phase 1 needed432 it).433434**Modern-web-guidance consult.** When the render implements435scroll-driven animation, view transitions, anchor positioning,436container queries, or perf-sensitive hero media, and the437`modern-web-guidance` plugin is installed, search it438(`npx -y modern-web-guidance@latest search "<query>"`) and follow439the retrieved guide; cite the guide id in440`_provenance.guidesConsulted[]`. Skip silently when the plugin is441absent.442443After craft returns, validate the output:444445- `:root` block present and complete (token-contract.md).446- Data attributes on every section (data-attributes.md).447- Anti-toolbox audit clean (each hit justified per divergence-toolkit.md448 § 1; record audit results in `DESIGN.json.extensions.divergence.anti_toolbox_hits`449 with the audit's amendments noted).450- Impeccable hard rules respected (OKLCH, type ratio ≥ 1.25, no451 reflex slop).452- **Content sourcing scan** — every literal value in the rendered453 output traces to one of the allowed sources454 (`reference/proposed-file-shell.md` § Content sourcing hierarchy).455 Any value that doesn't is either wrapped in a `[data-placeholder]`456 element with the mandatory visual signature, or the validation457 fails. Build the `_provenance.unsourcedContent[]` list during458 this scan.459460If validation fails, do not write the file. Surface the failure to461the user with the specific rule violated and a suggested fix.462463#### Craft-time disciplines (pre-write validators)464465Four disciplines fire on the rendered file *before* it lands on466disk. These run after craft returns its output and before the file467is written; failure refuses the write with a substitute proposal468(Discipline 6) or a rule citation (7, 8), and Discipline 9 registers469detector ignores rather than refusing.470471**Discipline 6 — Reflex-reject font pre-flight.** Grep the472declared `font-family` declarations against the reject list in473`skills/stardust/reference/divergence-toolkit.md` § Reflex-reject474fonts. If `≥ 3` reject-list families hit, refuse to write and475propose dimensionally-equivalent off-list substitutes from the476substitute table.477478The check fires only on rendered files whose `font-family`479declarations the agent had freedom to pick. **Mode A renders with480captured display + body families pinned (per481`direct/SKILL.md` § Mode A — Brand-faithful mode) bypass the482check** — the inherited families are not a reflex choice. Record483the bypass reason in `_provenance.reflexRejectAudit.bypassed` with484the captured families named.485486**Discipline 7 — Variable-font axis engagement.** When the resolved487direction's `expressive` or `distinctiveness` axes have moved past488their default (per489`skills/stardust/reference/intent-dimensions.md` §§ 2 + 5), engage490≥ 1 variable-font axis non-trivially per page using deck recipes491from the divergence toolkit. The toolkit's font decks expose named492axis recipes:493494- `serif-luxury` deck → Fraunces axis recipe: `opsz 144, SOFT 100, WONK 1`495- `bauhaus-functional` deck → Bricolage axis recipe: `opsz 96, wght 600`496- `tactile-humanist` deck → Recursive axis recipe: `MONO 1, CRSV 0, CASL 0.5`497498Static-weight static-style across all type on a page that claims499expressive or distinctive movement fails the check.500501**Static-only-family precedence (friction carve-out #1).** Family502substitution rules when the captured/pinned display font does not503ship a variable axis:5045051. If the captured/pinned display font is variable, engage its506 axes per deck recipes (standard path).5072. If the captured display font is **static-only by family**508 (Bellfort, GT Sectra static, hand-lettered custom cuts whose509 `woff2` files don't expose a parseable `fvar` table without510 runtime inspection), **exempt the display family** from the511 check and engage the **body family's axes** instead (typically512 `wght` and `ital` on a workhorse like Public Sans or Hanken513 Grotesk).5143. **Document** the family's static-only status in515 `DESIGN.json.extensions.divergence.font_deck.notes` so downstream516 pages don't re-run the check.517518A pre-flight that blindly grep'd for `font-variation-settings`519would fire on every brand-faithful render of a static-only-display520brand. The exemption keeps the spirit of the discipline (engage521variable behavior where it exists) while honoring the letter when522the captured family literally has no axes to engage.523524Off-deck accent fonts must ship with named **expressive positions**525(`marginal-tag`, `pointer-scribble`, `headline-callout`,526`badge-fill` — see toolkit § Font deck — expressive positions).527A face introduced without a position is a reflex pick, not an528expressive position.529530**Discipline 8 — Fidelity tier (`--fidelity=quick|refined|production`).**531532- `quick` (default) — sketch fidelity. Brief decisions land;533 micro-decisions stay provisional.534- `refined` — adds a craft micro-pass per535 `reference/fidelity-refined-pass.md`: formal type scale as CSS536 custom properties, `font-variant-numeric: tabular-nums` on inline537 digits, `text-wrap: balance` on headings + `text-wrap: pretty`538 on body + `hanging-punctuation: first` on `html`, sliding539 left-rule on nav hover, hairline hover on list items,540 kicker + title + baseline-aligned more-link section-head triplet,541 italic display couplet replacing structurally-vague `<b>`,542 bottom-of-sidebar reworked as typographic plate.543- `production` — `refined` + WCAG AA audit on every render + harden544 pass (loading states, error states, content-overflow edges).545546The tier is declared in the run invocation; persisted in547`_provenance.fidelity`. Default is `quick`.548549**Discipline 9 — Copy-cadence detector bypass under verbatim550fidelity.** This extends the Mode-A reasoning of Discipline 6 from551fonts to prose. impeccable's design detector ships prose-voice rules552(`em-dash-overuse`, `marketing-buzzword`, and similar copy-cadence553checks) that assume the copy is the agent's to rewrite. Under554`ia-fidelity: verbatim` — or any faithful/Mode-A render where the body555copy is `captured-verbatim` — that assumption is false by556construction: the prose is the source brand's, reproduced exactly per557the content-sourcing hierarchy, and rewriting it to satisfy a cadence558rule *is* the fabrication the fidelity setting exists to prevent. So559when the rendered file's copy classification is `captured-verbatim`560(per Discipline 5's `voiceClassification`), register those copy-cadence561rules as intentional ignores for the `<slug>-proposed.html` files562before the design hook fires — the same way Discipline 6 bypasses the563font reflex-reject check for pinned families. Scope the ignore to the564proposed files only, **never** to the project's own source (blocks,565styles, components), where the rules still apply because that copy566*is* the agent's. Record the bypass in567`_provenance.copyCadenceBypass` with the rules ignored and the568classification basis. The 2026-06-26 knack.com run hit this: the hook569flagged em-dashes and "enterprise-grade" on Knack's own headings570("Built on Enterprise-Grade Components") under a verbatim direction,571and the only correct response was to leave the captured copy untouched572and record the bypass.573574Bound the bypass: it covers prose-cadence rules only. Structural and575craft detector rules (`design-system-radius`, contrast failures,576reflex layout slop) are *not* exempted by verbatim fidelity — those577govern the agent's own CSS and structure, which faithful mode does not578freeze. Listing a rule under this bypass requires it be a copy-voice579rule whose subject is the captured prose.580581### Phase 2.4 — Motion application (when `--cinematic`)582583Fires only when `--cinematic` (with or without an explicit584register) was passed, OR when585`DESIGN.json.extensions.motion.register` was authored by `direct`586and the user did not opt out. Produces587`stardust/prototypes/<slug>-cinematic.html` **alongside** the static588`<slug>-proposed.html` — the static prototype is never replaced.589590Procedure:5915921. **Resolve the register.** For single-variant runs, read593 `DESIGN.json.extensions.motion.register`. **For multi-variant594 runs** (when `DESIGN-<id>.json` files exist at the project595 root), read `DESIGN-<id>.json.extensions.motion.register` for596 the variant currently being rendered — each variant may carry597 its own register, or omit it entirely (variant renders598 static). If `--cinematic=<register>` was passed at the CLI,599 the CLI value wins **only for the variants in scope** (every600 variant when no `<slug>` filter is set, otherwise the601 variant(s) matching the filter).602603 Record the source in `_provenance.motion.registerSource`604 (`"direct"` for per-variant DESIGN-authored, `"user-override"`605 for CLI). If neither path resolved a register for this606 variant, fall through to the selection heuristic in607 `reference/motion-registers.md` § Selection heuristic. If the608 heuristic itself returns no register (e.g. the variant's609 PRODUCT.md Brand Personality maps to none of the five610 registers, and no evidence-shaped extension register applies611 per the bank's § Extension rule in612 `reference/motion-registers.md`), skip Phase 2.4 entirely for613 that variant — render it static.614615 The per-variant resolution is what lets `uplift` produce a616 three-variant set where only variant C engages motion: `direct`617 writes the register into `DESIGN-C.json` only, and Phase 2.4618 fires for C alone (A and B render static).6196202. **Stage Lenis assets.** Copy621 `skills/prototype/assets/motion/lenis.min.js` and `lenis.min.css`622 into `stardust/prototypes/` (idempotent — skip if shas match).623 Cinematic prototypes load these via relative paths.6246253. **Read the canonical runtime.** Embed the inline script from626 `reference/motion-runtime.md` § The canonical script verbatim,627 with the `animConfig` constants rewritten per the active628 register's token defaults (`reference/motion-registers.md` §629 The five registers § Token defaults).6306314. **Layer the register's CSS.** Append the register-specific632 keyframes and class rules (entrance keyframes, parallax CSS633 custom properties, marquee animations, pulse animations) to634 the file's `<style>` block. The set of keyframes is closed635 per register; see `reference/motion-runtime.md` § Per-register636 tuning.6376385. **Annotate target HTML.** Walk the rendered DOM and emit the639 motion `data-*` attributes per the register's vocabulary:640 - `arrival`: `[data-anim]` on section heads, body copy,641 CTAs, and tile cards; `[data-countup]` on numeric values;642 `[data-parallax]` on the hero photograph.643 - `kinetic-display`: `[data-anim]` on most sections; `[data-split]`644 on display-cap headlines (`<h1 data-split>DINE</h1>`);645 `[data-flip]` on terminal codes / gate numbers; `.word`646 spans on display headlines for clip-path word wipes.647 - `live-systems`: `[data-tile-anim]` on every ops-tile and648 card; `[data-countup]` on every numeric data value;649 `[data-fill]` on every bar inner element; `.live-sweep`650 on the live-data container; `.marquee__track` on the top651 ticker.652 - `editorial`: `[data-anim]` only; `[data-parallax]` with653 reduced magnitude on hero imagery; never `[data-flip]` /654 `[data-fill]` / `[data-split]`.655 - `kinetic-grid`: `[data-tile-anim]` on cards; `[data-anim]`656 on section heads.657658 Full per-register attribute matrix: `reference/motion-registers.md`659 § Data-attributes consumed.6606616. **Inject the `<noscript>` fallback.** Add the no-JS override662 from `reference/motion-runtime.md` § No-JS fallback to `<head>`663 so the file degrades to its static-end state without664 JavaScript.6656667. **Update `_provenance`.** Add the `motion` block:667668 ```json669 "motion": {670 "register": "<register-name>",671 "registerSource": "direct | user-override | heuristic | extension",672 "runtimeVersion": "v1",673 "lenisAssets": { "js": "lenis.min.js", "css": "lenis.min.css" },674 "attributesEmitted": ["data-anim", "data-countup", ...]675 }676 ```6776788. **Hand off to Phase 2.8** (motion validation). The cinematic679 gates in `reference/motion-validation.md` § Pass 6 fire680 automatically because the rendered file declares the681 `_provenance.motion.register` field.682683The static `<slug>-proposed.html` is unaffected by this phase.684Both files are reviewable; the brand owner sees the cinematic685version when motion is part of the redesign brief, the static686version when migration / accessibility audit is the focus.687688#### Output paths689690| File | Owner phase | When |691|-------------------------------------|----------------------|-----------------|692| `stardust/prototypes/<slug>-proposed.html` | Phase 2 (always) | Always written. |693| `stardust/prototypes/<slug>-cinematic.html` | Phase 2.4 (`--cinematic`) | Written alongside; static remains. |694| `stardust/prototypes/lenis.min.js` | Phase 2.4 (`--cinematic`) | Copied from skill assets. |695| `stardust/prototypes/lenis.min.css` | Phase 2.4 (`--cinematic`) | Copied from skill assets. |696697#### When to use the static-only path698699The static prototype remains the load-bearing artifact for:700- Brand-faithful inheritance reviews (motion is additive — the701 static prototype is the canonical "yes, that's us, refreshed"702 surface).703- Accessibility audits (motion-driven pages are harder to evaluate704 in their reduced-motion state).705- Migration consumption (`migrate` reads the static prototype as706 its primary source. It does **not** merge the cinematic layer —707 per `skills/migrate/SKILL.md` § Phase 2 → Cinematic sibling, it708 carries the motion assets (`lenis.min.*`) through to709 `migrated/assets/motion/` and records710 `cinematic-variant-not-consumed` in the sidecar).711712The static prototype must pass every gate independently — the713cinematic layer cannot rescue a static prototype that fails714Phases 2.5–2.7.715716### Phases 2.5 – 2.8 — Quality gates: Critique → Audit → Adapt → Motion (Discipline 9)717718Four mandatory gate phases run by default before any prototype719can advance to `prototyped`. They implement Discipline 9: critique720covers *design*, audit covers *technical correctness*, adapt721covers *viewport behaviour*, motion covers *scroll-driven and722time-driven choreography correctness*. P0/P1 findings from **any**723of the four block `prototyped` until acknowledged. None have an724opt-out flag (per § No opt-outs).725726| Sub-phase | Focus | Catches |727|---|---|---|728| **2.5 Critique** | Design judgment | AI-slop reflexes, hierarchy regressions, contrast / cognitive issues, register drift |729| **2.6 Audit** | Technical correctness | a11y (alt, focus, contrast ratios computed), responsive overflow at 4–6 viewports, performance (LCP, image weights), JS-dependent-hidden-state |730| **2.7 Adapt** | Viewport behaviour | doc-width · overflow · sticky · grid columns · font scaling at 1920 / 1440 / 1280 / 800 / 414 / 375, mobile-nav-collapse audit |731| **2.8 Motion** | Scroll / time-driven correctness | clipped-container reveal timing, animation-range vs reading position, anim-enter trigger reachability, reduced-motion override completeness, no-JS fallback, multi-viewport scroll-driven check |732733Phase 2.8 fires only when the rendered file declares ≥ 1 named734choreography (per the page-shape brief's motion stack) OR uses735`animation-timeline:`, `@scroll-timeline`, IntersectionObserver-driven736entry triggers, or rAF loops reading `getBoundingClientRect()`,737**OR** when Phase 2.4 (motion application under `--cinematic`)738emitted a `<slug>-cinematic.html` file with motion attributes per739`reference/motion-attributes.md`. Cinematic prototypes additionally740trigger the cinematic-mode gates in `reference/motion-validation.md`741§ Pass 6 (Lenis bootstrap, reduced-motion fallback completeness,742scroll-jack check, three-position screenshots, register-match743audit, motion C-cliff detector). Static prototypes skip the744discipline entirely.745746The wasatch dry-run on 2026-05-13 caught two real bugs the brief747and craft phases missed (a WCAG miscalculation off by 0.27–1.86748points on multiple pairs of /beers variant C, and an LCP image749lazy-loaded on the first catalog entry). Both were caught by the750**audit** half — not by critique. The detector's contrast751computation and responsive performance check are the load-bearing752audits; without them the file would have shipped with quantifiable753WCAG failures.754755### Phase 2.5 — Critique756757Before opening the proposed file in the browser, run **two parallel758validators** against the rendered proposed file: `critique` and `audit`. They759are explicitly designed as a complementary pair — critique760covers *design* (AI-slop reflexes, hierarchy, brand fit,761cognitive load); audit covers *technical correctness*762(accessibility / performance / theming / responsive /763anti-patterns). Running only critique misses every quantifiable764WCAG / perf / responsive failure; running only audit misses765brand-misalignment and design slop. The pass is a **contract**,766not a courtesy.767768The 2026-05-04 nvidia.com home prototype critique returned7691 P0 + 2 P1 + 3 P2; the audit on the same artifact returned770**six additional findings** (no skip-link, theme carousels771without keyboard arrow nav, hero ~3.5MB without responsive772`<picture>`, layout-property animation, JS-gated reveal with773no `<noscript>` fallback, `scroll-behavior: smooth` not774respecting `prefers-reduced-motion`). None were design issues,775none would have been caught by critique alone. Without an776audit gate the page would have been marked `prototyped` with777quantifiable WCAG failures.778779Procedure:7807811. **Run both validators in parallel.** Invoke `impeccable:impeccable`782 twice in the same Skill-tool batch:783784 ```785 Skill { skill: "impeccable:impeccable",786 args: "critique stardust/prototypes/<slug>-proposed.html --json" }787788 Skill { skill: "impeccable:impeccable",789 args: "audit stardust/prototypes/<slug>-proposed.html --json" }790 ```791792 Each returns a JSON findings list — each finding has793 `priority` (P0 / P1 / P2 / P3), `category` (hierarchy /794 contrast / motion / a11y / perf / responsive / etc.), and a795 one-line description. Capture critique findings into796 `_provenance.critique[]` and audit findings into797 `_provenance.audit[]` on the proposed file (append; never798 overwrite previous runs' entries).7998002. **Brand-faithful inversion auto-dismiss.** Both validators801 ship known false positives on Mode A renders — Arial fallback802 reads as "overused-font," eyebrow uppercase reads as803 "all-caps body," pure white / pure black flagged when the804 brand's captured palette includes them. Before surfacing805 findings to the user, diff each finding against806 `DESIGN.json#extensions.divergence.brand_faithful_inversions[]`807 and `DESIGN.md#narrative.rules` (e.g. permitted uppercase808 contexts). Drop findings whose category and target match an809 approved inversion; keep the original list in810 `_provenance.<critique|audit>[]` with a811 `dismissedAsBrandFaithful: true` flag for audit-trail812 purposes. The user-facing report shows only the real hits.8138143. **Vision gate.** Render a screenshot of the proposed file and815 study it NEXT TO the captured source screenshot816 (`stardust/current/assets/screenshots/<slug>.png` when present).817 Judge visually, not from the DOM: **brand-fit** (would the818 brand owner say "that's us"?), **signature preservation** (hero819 medium / motif carried, per Discipline 3's signature clause),820 and **hierarchy at a glance** (does the eye land where the821 brief says it should?). Record822 `_provenance.visionCheck = { verdict: pass|fail, observations[] }`;823 a `fail` is a P1 finding through the same gate mechanics as824 critique/audit findings. The vision gate complements — never825 replaces — the deterministic critique/audit pair.8268274. **Surface findings in the user-facing report**, grouped by828 priority across both validators with the source attributed829 (`critique:` / `audit:`). List the first 5 P0/P1 verbatim;830 collapse P2/P3 to per-source counts with an "expand to see831 all" pointer. Format:832833 ```834 Critique + audit on home-proposed.html835836 P0 (1)837 audit: skip-link missing — header has no <a href="#main"> as first focusable838 P1 (4)839 critique: hierarchy regression — H2 visually heavier than H1 in section #features840 audit: hero <img> 3.5MB; no responsive <picture> set despite captured srcset841 audit: transition: width animates layout properties (jank)842 audit: .hero scroll-behavior: smooth not gated by prefers-reduced-motion843 P2 (3 critique, 1 audit) — expand to see844 P3 (0)845 ```8468475. **Gate `prototyped` status on P0/P1 findings from EITHER848 validator.** If the merged-and-deduped findings list (after849 the brand-faithful auto-dismiss, plus a vision-gate `fail`850 from step 3, which counts as P1 here) contains any P0 or P1,851 do **not** mark the page `prototyped` in `state.json` yet. The852 proposed file is on disk and openable in the browser, but the853 page stays in `directed` until either:854 - The agent fixes the issue (run a chat-driven impeccable855 command per Phase 4 iteration paths, then re-run Phase 2.5).856 - The user explicitly acknowledges (e.g. "ship as-is" /857 "accept the P1 findings"). Acknowledgement is recorded in858 `_provenance.critique[]` AND `_provenance.audit[]` with the859 verbatim user phrase, so re-runs see the existing860 acknowledgement and don't re-prompt.861862 P2/P3 findings do not block `prototyped`. They surface as863 advisory.8648656. **Optionally spawn an LLM design-review subagent** for an866 independent take when the user wants more than the867 deterministic detector. Trigger only when the user explicitly868 asks ("give me a deeper critique", "second opinion") or when869 the deterministic pass returns ≥3 P0/P1 findings (signal that870 the render has multiple issues worth a closer look). Default871 off to keep the loop fast.872873Both validators are mandatory. There is no opt-out flag — if a874finding bites and the user wants to ship anyway, they say so in875chat ("ship as-is" / "accept the P1 findings") and the876acknowledgement is recorded in `_provenance.critique[]` AND877`_provenance.audit[]` verbatim. The contract is that the gate878runs; the user can override the gate but the override is879explicit and recorded.880881Failure handling: when impeccable is unavailable per the882Delegation mechanic, prototype refuses to run (impeccable is a883hard requirement). There is no degraded mode that ships884unverified output.885886### Phase 2.6 — Audit (detector specifics)887888The audit sub-phase is implemented as the second parallel arm of889the impeccable invocation in § Phase 2.5 above. This section890documents the audit-specific detectors that fire in that arm,891including ones the parallel critique pass does not cover.892893**WCAG contrast computation.** Audit recomputes contrast ratios894for every text-on-substrate pair in the file. The previous895brief-time / craft-time hand-estimated `_provenance.wcagContrast`896blocks are NOT authoritative; the audit's computed values are.897Discrepancies of `≥ 0.25` between hand-estimated and computed898ratios surface as a P1 finding. The wasatch dry-run on 2026-05-13899surfaced 0.27–1.86 point discrepancies on /beers variant C; the900audit-driven fix darkened the Cutthroat substrate from `#1c8a7a`901to `#177566` to clear AA body (4.5:1) and recolored the902pour-link from yellow to white.903904**Large-text exemption check.** A pair claimed to qualify for905WCAG's "large-text" 3:1 floor (instead of body 4.5:1) must be906`≥ 18pt regular` OR `≥ 14pt bold` at rendered size. The audit907verifies the size at the rendered DOM, not at the source CSS. A90813px Bellfort 700 link claiming the large-text exemption fails909the check (Bellfort 700 at 13px is below the 14pt bold floor); the910finding is P0.911912**LCP image audit.** The first image above the fold on the page913must declare `loading="eager"` AND `fetchpriority="high"`. A914lazy-loaded LCP image is a P1 finding. Detection: the first `<img>`915whose computed `getBoundingClientRect()` intersects the initial916viewport must satisfy both attributes.917918**JS-dependent-hidden-state detector.** Per919`notes/prototype-broken-by-default-detector-2026-04-29.md`, the920audit catches initial-state CSS that hides content via `clip-path:921inset(0 100% ...)`, `opacity: 0`, or `transform: translateX(-100%)`922where the reveal depends on a JS class flip (typically an923`IntersectionObserver` toggling `.in-view` / `.is-visible` /924`html.js-anim`). When the observer fails (slow JS, blocked CDN,925NoScript), the element stays permanently invisible.926927Detection regex over the CSS text:928929```930/(clip-path:\s*inset\(0\s+100%|opacity:\s*0\s*[;}]|transform:\s*translate[XY]\(-?100%)/931```932933When a hit is found AND no `<noscript>` fallback styles the same934selectors visible AND no `prefers-reduced-motion` rule short-935circuits the hide, the finding is P0 (the page is broken-by-default936for any user whose JS doesn't execute). Mitigations the page may937declare in `_provenance.audit.acknowledged[]`:938939- A `<noscript>` block that re-sets the hidden selectors to940 visible.941- A `prefers-reduced-motion: reduce` block that skips the hide942 entirely.943- An animation gate selector pattern (`html.js-anim .selector { ...944 hidden ... }`) where the gate class is set inline in `<head>`945 BEFORE the stylesheet loads, so observer failure doesn't strand946 elements (the wasatch /beers variant C confirms this pattern947 works — detector PASS on all 4 files via the `html.js-anim` gate948 + CDN-failure fallback).949950### Phase 2.7 — Adapt (mandatory pre-approval)951952Every prototype goes through `$impeccable adapt` and the adapt953audit before the user's approval is accepted. The cascade ships954desktop-only HTML otherwise — viewports are tuned to ~1440×900955through render and iteration, and nothing earlier in the pipeline956produces responsive coverage.957958This phase was Phase 5.5 in prior versions of stardust (running959*after* approval). The current spec promotes it to Phase 2.7 so it960gates `prototyped` alongside critique and audit. The substantive961adapt logic is unchanged; only the trigger point moved (pre-962approval vs post-approval). See § Adapt procedure below for the963unchanged implementation.964965#### Adapt procedure966967Invoke impeccable adapt against the rendered file:968969```970Skill {971 skill: "impeccable:impeccable",972 args: "adapt stardust/prototypes/<slug>-proposed.html"973}974```975976Pass the captured viewport breakpoints from977`DESIGN.json#extensions.breakpoints` if present; otherwise adapt978picks defaults (640 / 768 / 1024 / 1280 are the stardust spec979defaults — anything above 640 used as a "mobile breakpoint" is a980smell per § Mobile-adapt audit below). The full target list is981**1920 / 1440 / 1280 / 800 / 414 / 375** — adapt validates each982viewport's doc-width · overflow · sticky behaviour · grid columns983· font scaling.984985Validate the adapted file against the same contracts Phase 2 ran986(`:root` token block, data attributes, anti-toolbox audit clean,987impeccable hard rules, content sourcing). Adapt is an iteration988over the existing render; it must not reintroduce contract989violations.990991Append an entry to the proposed file's `_provenance.adapt[]`992recording: ISO timestamp, the breakpoint list applied, the number993of `@media` rules added, and any layout decisions the adapt pass994surfaced (carousel → stack, sidebar → drawer, hamburger → menu).995996#### Mobile-adapt audit997998Phase 2.7 also runs a hard audit, separate from the adapt pass999itself, that the resulting file would survive a publish or1000migrate. The same audit re-runs at `migrate` and `--publish-sample`1001so an adapted-but-broken render can't slip through.10021003Refuse the file when **any** of:10041005- `<meta name="viewport" content="width=device-width, ...">` is1006 missing or width is set to a fixed pixel value.1007- The file declares zero `@media (max-width: ...)` rules at all1008 (desktop-only).1009- The file declares mobile-targeted breakpoints **above 640px**1010 — `@media (max-width: 1024px)` as the *narrowest* breakpoint1011 is the recognisable shape of "didn't actually adapt for1012 phones." Adapt should produce at least one rule at ≤ 640px.1013- **At a 360px-wide rendered viewport:** a landmark causes1014 `scrollWidth > clientWidth` on `document.documentElement` or1015 `document.body`. Refusal code:1016 `audit/responsive: horizontal-overflow-at-360px`.1017- **At a 360px-wide rendered viewport:** the computed `font-size`1018 of any descendant of a `<nav>` inside a `<header>` is below1019 11px. Refusal code: `audit/responsive: nav-readability-floor`.1020- **At a 360px-wide rendered viewport:** the computed `gap` (or1021 `column-gap`) of any flex/grid `<nav>` inside a `<header>` is1022 below 10px. Same refusal code.10231024The last three conditions require actually rendering the file —1025they can't be inferred from CSS text. The canonical check is a1026small Playwright snippet (`fixtures/mobile-nav-audit.mjs`) the1027agent runs against the file at 360×800. Audit messages must1028include a pointer to the stock-template doc:10291030> Suggested fix: apply the stock hamburger pattern1031> skills/prototype/reference/mobile-nav-collapse.md10321033#### Mobile nav collapse10341035When the audit refuses on either of the nav-related codes above,1036the agent applies the stock CSS-only hamburger pattern documented1037in `reference/mobile-nav-collapse.md` (HTML + CSS + ≤10-line1038inline `<script>` for a11y). The header gains1039`data-nav-collapse="hamburger"` so downstream consumers can detect1040the pattern is applied.10411042**Source order is load-bearing.** The base1043`.ds-nav-burger { display: none; ... }` rule must be placed at1044the **top** of the `<style>` block (after `:root`, before any1045`@media` rules). The post-injection ordering check is a small1046`awk` one-liner documented in `reference/mobile-nav-collapse.md`1047§ Source order — it asserts the base rule appears at a lower line1048number than the first `@media (max-width: 640px)` rule. Exit 0 =1049correct ordering; exit 1 = regression.10501051The stock pattern is the default; `reference/mobile-nav-collapse.md`1052§ Alternative patterns documents priority+overflow, bottom nav,1053and side-drawer as valid alternatives the user can request, but1054the agent does not pick between them autonomously.10551056### Phase 2.8 — Motion validation (mandatory when motion declared)10571058Static-DOM gates (2.5 critique, 2.6 audit, 2.7 adapt) all read the1059proposed file at `scrollY=0` with no time elapsed. They cannot1060catch motion-specific failure modes:10611062- Content stuck `opacity: 0` because an anim-enter trigger fires1063 past the user's reading position.1064- Content rendered outside its parent's `overflow: clip` boundary1065 because a translateY animation magnitude exceeds the parent's1066 slack room.1067- Scroll-driven animations whose `animation-range` completes only1068 after the section has fully scrolled past, leaving the reveal1069 state invisible during the readable window.1070- rAF loops with stale baseline measurements drifting on resize /1071 font-load / lazy-image-load.1072- Section overlaps caused by transforms that don't appear in1073 static DOM.1074- `prefers-reduced-motion` regressions where the choreography is1075 disabled but the element is left at the hidden "from" state.1076- No-JS regressions where the choreography's hidden initial state1077 has no `<noscript>` fallback.10781079The full procedure (5 passes — scroll-position probe, finding1080classification, motion-bug checks, no-JS state, multi-viewport1081scroll-driven check) lives in `reference/motion-validation.md`.10821083Phase 2.8 fires when the rendered file declares ≥ 1 named1084choreography (per `_provenance.motion.choreographies[]`) OR contains1085any of: `animation-timeline:`, `@scroll-timeline`, IntersectionObserver1086with `target.classList.add('anim-enter-visible')` patterns, or rAF1087loops driven by `getBoundingClientRect()`. Static prototypes skip.10881089Findings are classified as **by-design** (the choreography is1090supposed to produce this state at this position; explained by a1091named choreography in the brief) or **bug** (the choreography1092produces an unintended state). Bugs block `prototyped` until1093acknowledged or fixed.10941095When a bug cannot be auto-fixed within 3 iterations of the recursive1096loop, surface to the user with the finding, the classification1097reasoning, the 3 fix attempts that were tried, and a proposed1098remediation that requires user input. Do not silently lower the1099gate.11001101Append an entry to the proposed file's `_provenance.motionValidation`1102recording: ISO timestamp, probe positions run, viewports tested,1103findings (hiddenInViewport, sectionOverlaps, clippedReveals,1104rangeMismatches) with per-finding classification, and `fixesApplied[]`.1105Save clean-pass screenshots to `stardust/validation/<slug>/motion-<viewport>.png`.11061107#### Variant-convergence detector (Discipline 10)11081109When N > 1 variants render, each `<slug>-<id>-shape.md` declares:11101111- A distinct `dominantDimension` value (no two variants share it;1112 per `notes/variant-convergence.md` Tier 1).1113- A `compositionDelta` field listing ≥ 2 ways the variant's section1114 sequence or layout strategy diverges from each sibling variant.1115 Examples:1116 - `compositionDelta: ["section-order: hero ↔ stats", "layout: 3-up-grid → vertical-narrative"]`1117 - `compositionDelta: ["ia-priority: commercial-conversion → catalog-browse", "hero-layout: split-photo → type-led"]`11181119If `compositionDelta` is empty or trivial (only token-level1120deltas — *"Color A vs Color B"*, *"font weight 400 vs 600"*),1121Phase 1 fails the brief and restarts ideation. The validator1122checks pairwise: variant pairs `(A,B)`, `(A,C)`, `(B,C)` each1123need ≥ 2 structural changes.11241125**Mode-aware contract** (composes with `ia-fidelity` from1126`stardust/reference/intent-dimensions.md` § 9):11271128- Under `ia-fidelity: verbatim` — A1/A2/A3 differ on **surface**1129 axes only (type-weight, type-scale, density, motion-energy,1130 color-temperature, spacing-rhythm). Structural deltas (section1131 sequence / presence / IA priority / layout strategy) are1132 **forbidden** under verbatim. The detector inverts: a surface-1133 only delta passes, a structural delta refuses.1134- Under `ia-fidelity: reimagined` — A + B + C variants declare1135 structural `compositionDelta` with ≥ 2 changes per pair.1136 Surface-only deltas **fail** the brief.11371138The wasatch /beers A / B / C dry-run declared 5 structural deltas1139per pair (`compositionDelta_vs_A`, `compositionDelta_vs_B` blocks1140in `_provenance`) — substrate-strategy, section-presence, photo-1141treatment, section-order, section-completeness. The detector1142PASSED on all three pairs.11431144Single-variant runs (N = 1) skip Discipline 10 entirely.11451146### Phase 4 — Open and iterate11471148(Phase numbering skips 3; the former Phase 3 — *Compose the viewer*1149— was removed when the per-page before/after viewer was dropped.1150Cross-references throughout the docs still name Phases 4, 5, 5.5.)115111521. Open the just-written or just-updated `<slug>-proposed.html` (or1153 the user-chosen variant suffix when N > 1) in the browser using1154 the `open <vfs-path>` shell command. This routes VFS paths through1155 the preview service worker — **do not use `playwright-cli open`**1156 for local prototype files, as it bypasses the preview service worker1157 and produces a FILE NOT FOUND error. Skip in pipeline-automation mode.11582. Mark the page `prototyped` in `state.json` — **gated on the1159 Phase 2.5 critique + Phase 2.6 audit + Phase 2.7 adapt +1160 Phase 2.8 motion validation (when fired) result** (Discipline 9).1161 If any of the four returned ≥ 1 P0 or P1 finding (after the1162 brand-faithful inversion auto-dismiss) and the user has not1163 acknowledged, the page stays `directed` (not `prototyped`);1164 surface the findings in the report grouped by source1165 (`critique:` / `audit:` / `adapt:` / `motion:`) and recommend1166 either fixing the issue or acknowledging explicitly. The1167 transition itself does not require *approval* (a separate1168 later step) — but it does require all gates to clear, since1169 shipping a `prototyped` flag on work that fails P0/P1 on any1170 gate misleads downstream consumers (migrate, the dashboard)1171 about the prototype's quality.11723. Report the prototype path and stop. Iteration happens via1173 chat-driven impeccable commands or direct invocation (see1174 § Iteration paths below).11751176#### Iteration paths11771178Refinement after the initial render takes one of two forms. They1179are not mutually exclusive — a single page can move through both1180across its lifetime.118111821. **Chat-driven (default).** The user gives a refinement phrase1183 in chat — *"make the hero bolder for home"*, *"tighten the1184 cup-note grid"*, *"less corporate"*. The agent:1185 - Reads the phrase against1186 `skills/stardust/reference/intent-dimensions.md` to identify1187 which axes it moves.1188 - Consults1189 `skills/stardust/reference/impeccable-command-map.md` to pick1190 the matching impeccable command (often `bolder`, `quieter`,1191 `distill`, `typeset`, `colorize`, or `layout`).1192 - Shows the resolved plan to the user before executing.1193 - Runs the chosen command against `<slug>-proposed.html` (or a1194 specific section within it, when the phrase scopes one).1195 - Re-validates per Phase 2 (`:root` block, data attributes,1196 anti-toolbox audit clean, impeccable hard rules) and updates1197 the proposed file's provenance.119811992. **Direct impeccable invocation.** The user runs an impeccable1200 command directly — `$impeccable bolder1201 stardust/prototypes/home-proposed.html`. Stardust isn't in the1202 loop; the browser tab reloads whatever's on disk. This is fine1203 and documented as a supported escape hatch. (Includes1204 `$impeccable live` against the proposed file when the user1205 wants in-browser picker iteration — that's an external tool1206 they invoke directly; stardust does not drive its poll loop.)12071208The "open and reasoned" principle from the master skill applies to1209path 1: the agent reasons publicly about the phrase before running1210any command, and never silently maps a refinement to a fixed1211command.12121213### Phase 5 — Approval (+ fold-back, Part III)12141215Approval is **explicit**. Stardust does not auto-approve.12161217The user signals approval by saying **"approve <slug>"** or simply1218**"approve"** in the conversation. The agent confirms the slug1219before writing state, then proceeds.12201221On approval:122212231. Verify the proposed file's provenance block lists the *current1224 active* `direction.md` (defensive check — if the direction changed1225 during iteration, the user must re-prototype against the new1226 direction first).12272. **Run approval fold-back** per1228 `reference/approval-fold-back.md` (Part III of the merged spec).1229 When the approved variant is non-A AND `ia-fidelity` is1230 `reimagined`, the prototype reads the approved file's1231 `_provenance.differentiationFromA[andB]` block plus the brief's1232 `compositionDelta`, diffs against the active direction, and1233 proposes folding any structural moves. The user picks1234 `fold site-wide` (default) / `fold page-local` / `don't fold`.1235 Under `ia-fidelity: verbatim` the fold-back is a no-op by1236 construction (A1/A2/A3 are surface forks; nothing structural to1237 fold). Flags: `--auto-fold` skips the gate; `--no-fold` opts out1238 entirely.12393. Mark the page `approved` in `state.json`. Append a1240 `{ status: "approved", at: <ts> }` history entry.12414. Clear any `stale` flag on the page.12425. Print:1243 ```1244 home: approved1245 proposed: stardust/prototypes/home-proposed.html1246 mobile: adapted at Phase 2.7 (4 @media rules at 640/768/1024/1280)1247 fold-back: page-local (substrate-keyed-zine pattern stays on /beers)12481249 Next: $stardust migrate home (write final redesigned static HTML)1250 ```12511252Adapt no longer runs at approval time — it ran at Phase 2.7 as a1253pre-`prototyped` gate. The `mobile:` line reports the Phase 2.71254result for review continuity.12551256If multiple pages are in flight, approval is per-page; the user can1257approve some and continue iterating on others.12581259### Phase 5.5 — Adapt for mobile (retired; see Phase 2.7)12601261Adapt and the mobile-adapt audit moved to **Phase 2.7** as part of1262the Discipline 9 critique → audit → adapt cycle, gating1263`prototyped` instead of approval. The 2026-05-03 lovesac.com1264showcase failure mode (stakeholders eyeballing variants on a phone1265getting the unadjusted desktop layout) is unchanged — the gate just1266fires earlier in the lifecycle, so a non-adapted prototype never1267lands as `prototyped` in the first place.12681269The implementation (impeccable adapt invocation, mobile-adapt1270audit, mobile-nav-collapse stock pattern) lives in § Phase 2.71271above. This section is retained as an anchor for any external1272cross-reference that still names Phase 5.5.12731274### Stale handling12751276When `direction.md` changes, the prototype's `againstDirection`1277provenance becomes outdated and `state.json` flags the page1278`stale: true`. Default behaviour:12791280- `$stardust prototype` (no slug) skips stale pages and reports the1281 count: `2 stale pages (home, about) — re-run with --all.`1282- `$stardust prototype home` operates on `home` even if stale.1283- `$stardust prototype --all` re-prototypes every directed page1284 including stale ones.12851286When a stale page is successfully re-prototyped, clear its `stale`1287flag and update `againstDirection` to the new active direction.12881289## Outputs12901291| Path | Purpose |1292|-----------------------------------------------|-----------------------------------------------|1293| `stardust/prototypes/<slug>-shape.md` | Per-page compositional brief (Phase 1 output, craft input). |1294| `stardust/prototypes/<slug>-proposed.html` | Proposed redesign (chat-driven iteration target, migration source, user-facing review surface). |1295| `stardust/state.json` | Updated with page status and approval history. |1296| `DESIGN.json` | Updated with `extensions.divergence.anti_toolbox_hits` and any audit amendments from this prototype's render. |12971298## Failure modes12991300- **No directed pages.** Recommend `$stardust direct` and stop.1301- **Pending direction.** Refuse to run; the user must resolve the1302 direction first.1303- **Validation failure (:root block missing, data attributes missing,1304 unjustified anti-toolbox hit, impeccable rule violation).** Do not1305 write the file. Surface the specific failure and a suggested fix.1306- **Impeccable not available.** Refuse to run — impeccable is a1307 hard requirement (Delegation mechanic). Recommend the user1308 install the impeccable plugin and re-invoke.13091310## Concurrency13111312Per `state-machine.md`: stardust does not lock. Two concurrent1313`prototype` runs on different slugs are safe. Two on the same slug1314are last-write-wins; warn the user if they explicitly try.13151316## Prep mode (--prep)13171318When invoked with `--prep`, prototype runs an extended pass that1319fills page-type gaps and writes canon. Discovery-mode runs are1320unchanged: per-slug shape brief, render via `$impeccable craft`,1321open in browser, iterate, approve.13221323`--prep` adds three things on top of the standard procedure:13241325### 1. Fill page-type gaps13261327Identify every page type in `state.json.pages[].type` that1328doesn't yet have an approved archetype. For each gap, prototype1329one representative page (the user picks which slug, or the first1330page of that type by default):13311332- `article`-typed pages with no approved article: prototype one1333- `listing`-typed pages with no approved listing: prototype one1334- `program`, `form`, `static` — same pattern1335- `landing` — the home; prototyped first if not already done1336- `unique`-typed pages — these don't get archetypes; they're1337 rendered as one-offs at migrate time13381339The user picks one variant per type. Subsequent pages of the1340same type are migrated by forking that approval (Path A′ in1341`skills/migrate/SKILL.md`).13421343### 2. Canon write-back on first approval13441345The first prototype approved (default the home; override with1346`--canon-from <slug>`) becomes the **canon-author**. On1347approval, extract canon and write back per1348`reference/canon-extraction.md`:134913501. Chrome HTML → `stardust/canon/header.html`,1351 `stardust/canon/footer.html`, optional regions.13522. Compound CSS → `stardust/canon/canon.css`.13533. Pinned tokens → `DESIGN.json.extensions.canon.pinned`.13544. Module canonical renderings →1355 `stardust/canon/modules/<id>.html`.13565. Compositional moves (LLM-authored, 3–7 lines) →1357 `DESIGN.json.extensions.canon.compositionalMoves`.13581359Reference all extracted files via `{ path, sha }` in1360`DESIGN.json.extensions.canon.files`. Each canon file carries a1361`stardust:canon` provenance comment naming source slug, source1362prototype, and region.13631364### 3. Canon write-back on subsequent approvals13651366For non-canon-author template approvals (article, listing, etc.),1367canon-extraction runs in **diff mode** per1368`reference/canon-extraction.md` § Conflict resolution:13691370- **Net-new items** (a module not yet in canon, a new compound1371 CSS class, a new compositional move) → append to canon1372 additively. Add a history entry naming what was added.1373- **Match canon byte-for-byte** → no-op.1374- **Conflict with canon** → default is **log as deviation**:1375 the migrated/prototyped page carries the deviation inline1376 marked with `data-deviation="<reason>"`, and the page's1377 `migrationDecisions[]` records a `canon-deviation` entry.1378 Canon stays unchanged.13791380Override the default per-conflict via the prep summary if the1381user wants to promote the new variant to canon (which stale-flags1382downstream pages that consumed the changed item) or reject and1383re-iterate the prototype. Without an explicit override, the1384conflict logs as deviation and approval proceeds.13851386### Prep summary13871388```1389prototype --prep complete1390=========================13911392Approved archetypes:1393 landing home (V01 Polish) canon-author1394 article news/post-housing-summit1395 listing news (the index)1396 program programs/shelter1397 form donate1398 static about13991400Canon: stardust/canon/ + DESIGN.json.extensions.canon1401 Sources: home → article (1 deviation logged), listing (clean),1402 program (1 deviation logged), form (clean), static (clean)1403 Modules: 8 confirmed; canonical renderings written1404 Pinned: sectionPadding, densityTier, typeScale set14051406Next: $stardust migrate (apply canon to every page in inventory)1407```14081409Default mode is unchanged.14101411## References14121413- `reference/page-shape-brief.md` — per-page compositional brief1414 format (Phase 1 output, craft input). Page-level deployment1415 decisions live here; site-level system decisions live in1416 DESIGN.md.1417- `reference/canon-extraction.md` — the five-step canon-extraction1418 procedure performed on prototype approval in `--prep` mode.1419- `reference/proposed-file-shell.md` — proposed-file schema and1420 required structure (`:root` token block, data attributes,1421 provenance, content sourcing hierarchy, mobile-adapt audit).1422- `reference/publish-sample.md` — `--publish-sample` sub-flow:1423 eligibility checks, file staging, PR creation against the1424 upstream stardust showcase. Lands the prototype as a public1425 sample at `https://{owner}.github.io/stardust-2/`.1426- `reference/mobile-nav-collapse.md` — stock CSS-only hamburger1427 pattern Phase 2.7 auto-applies when the Mobile-adapt audit1428 refuses on `horizontal-overflow-at-360px` or1429 `nav-readability-floor`. Carries the copy-pasteable1430 HTML+CSS+JS, the audit smoke-test command, and the1431 alternative-pattern vocabulary.1432- `reference/motion-validation.md` — Phase 2.8 procedure:1433 scroll-position probe, finding classification, motion-bug1434 checks (clipped-container reveal timing, animation-range vs1435 reading position, anim-enter trigger reachability,1436 reduced-motion override completeness, no-JS state, multi-1437 viewport scroll-driven check). Reusable Playwright probe1438 patterns documented inline.1439- `reference/motion-registers.md` — five brand-faithful motion1440 registers (`arrival`, `kinetic-display`, `live-systems`,1441 `editorial`, `kinetic-grid`), the selection heuristic that1442 maps PRODUCT.md Brand Personality traits to a register, and the1443 § Extension rule for bespoke registers derived from the site's1444 own captured motion. Consumed by Phase 2.4 (motion application).1445- `reference/motion-stack.md` — technology choice for cinematic1446 prototypes: Lenis + CSS keyframes + rAF + IntersectionObserver.1447 Why not GSAP. Bundle policy + Lenis pinning procedure.1448- `reference/motion-attributes.md` — `data-*` vocabulary the1449 cinematic runtime consumes (`[data-anim]`, `[data-tile-anim]`,1450 `[data-countup]`, `[data-flip]`, `[data-fill]`, `[data-split]`,1451 `[data-parallax]`). Annotated per the active register.1452- `reference/motion-runtime.md` — the canonical inline runtime1453 script that powers every cinematic prototype. Single source of1454 truth; embedded verbatim during Phase 2.4.1455- `reference/fidelity-refined-pass.md` — concrete CSS recipes for1456 the `--fidelity=refined` craft micro-pass (Discipline 8).1457- `reference/anti-template-bank.md` — worked examples of the1458 non-template moves Discipline 3 draws from (typographic1459 substitution / substrate-promotion / inversion / document-shape1460 / scale-displacement), plus the § Extension rule for1461 evidence-shaped new moves.1462- `skills/stardust/reference/reference-research.md` — sourcing1463 real-world design references (refero MCP with WebSearch fallback,1464 graceful skip when unavailable); consumed by Discipline 2's1465 reference-grounded alternatives.1466- `reference/approval-fold-back.md` — Phase 5 fold-back procedure1467 (Part III of the merged spec): diff algorithm, surfacing UX,1468 write logic, stale flagging, `--auto-fold` / `--no-fold` flags,1469 verbatim no-op gating.1470- `fixtures/composition-delta-good.md` — example shape brief with1471 a strong structural `compositionDelta` (passes Discipline 10).1472- `fixtures/composition-delta-trivial.md` — example shape brief1473 with token-only deltas (fails Discipline 10).1474- `../stardust/reference/token-contract.md` — `:root` token1475 block (cross-cutting, used by prototype + migrate).1476- `../stardust/reference/data-attributes.md` — structural data1477 attribute vocabulary (cross-cutting, used by prototype + migrate).1478- `../stardust/reference/divergence-toolkit.md` —1479 anti-mediocrity rules consumed during render and iteration.1480- `../stardust/reference/intent-dimensions.md` — the 7-axis1481 vocabulary used to read a chat-driven refinement phrase1482 (iteration path 2).1483- `../stardust/reference/impeccable-command-map.md` — when to1484 reach for each impeccable command. Consulted during chat-driven1485 iteration (path 2) to pick the command for a refinement phrase.1486- `../stardust/reference/state-machine.md` — page lifecycle1487 and stale rules.1488- `../stardust/reference/artifact-map.md` — provenance shape.1489- impeccable's `reference/craft.md` and `reference/live.md` — the1490 underlying impeccable commands stardust delegates to.