mirror of
https://github.com/adobe/skills.git
synced 2026-09-19 08:06:09 +08:00
content-driven-development-v2.0.1
18 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 David Bosschaert
|
ecd1e2b2c4 |
fix(stardust): fix opening prototyped pages from SLICC browser plugin (#152)
* fix(stardust): fix opening prototyped pages from browser plugin Use `open <vfs-path>`, not `playwright-cli open`, for prototype previews. VFS-backed prototype files must be opened with the `open` shell command so the preview service worker resolves the path. `playwright-cli open` bypasses the worker and yields FILE NOT FOUND. Update prototype and uplift SKILL.md to spell this out. * chore: remove multiple files detail |
||
|
Paolo Mottadelli
|
07aa1f4211 |
fix(stardust): close 3 Tessl orphaned-reference warnings (#151)
Closes three low-risk Tessl skill-review orphan warnings. - distill/SKILL.md: added a `## References` section listing the two reference files (samples-brief-format.md, trait-matrix-schema.md). distill was the only skill without a References section. - prototype/SKILL.md References: added the four cinematic motion docs (motion-registers.md, motion-stack.md, motion-attributes.md, motion-runtime.md) alongside motion-validation.md. - prototype/SKILL.md References: switched 7 cross-skill paths from `skills/stardust/reference/X.md` (absolute-style, unresolvable by Tessl) to `../stardust/reference/X.md` (the convention already used by stardust/SKILL.md and uplift/SKILL.md). No behavior changes — additive + path-correctness fixes. 🤖 Generated with [Claude Code](https://claude.com/claude-code) |
||
|
Paolo Mottadelli
|
07fa21891a |
feat(stardust): 0.7.1 → 0.10.0 — distill, cinematic motion, uplift orchestrator (#149)
Stacks three minor versions on the `redesign-adobecom` branch. ## 0.8.0 — Original `redesign-adobecom` changes - NEW skill: `distill` — analyze HTML mockups + website URLs and produce a structured style guide; anchors `direct` under Mode B. - Motion validation gate (Phase 2.8 on `prototype`) — clipped-container reveal timing, animation-range vs reading position, anim-enter trigger reachability, reduced-motion override completeness, no-JS fallback, multi-viewport scroll-driven checks. - Surface forks formalized as parent-direction tuning under reimagined mode. - Journal rule + recursive validate-and-fix loop promoted to master skill. - HTTP/2 / Akamai bypass: route-fulfiller pattern in playwright-recipe. ## 0.9.0 — Cinematic motion feature - `prototype --cinematic` (and `--cinematic=<register>`) — layers a brand-faithful motion register on the static prototype. Output is `<slug>-cinematic.html` alongside the static `<slug>-proposed.html`. - Five motion registers as a closed catalog (`arrival`, `kinetic-display`, `live-systems`, `editorial`, `kinetic-grid`). - Lenis bundled at `skills/prototype/assets/motion/` (MIT, ~17 KB). No GSAP / Framer / Theatre.js. - Motion `data-*` vocabulary: `[data-anim]`, `[data-tile-anim]`, `[data-countup]`, `[data-flip]`, `[data-fill]`, `[data-split]`, `[data-parallax]`. - Canonical inline runtime at `prototype/reference/motion-runtime.md`. - Six cinematic-mode validation gates (Pass 6): Lenis bootstrap, reduced-motion fallback, scroll-jack check, three-position screenshot pass, register-match audit, motion C-cliff detector. - `direct` writes `extensions.motion.register` from PRODUCT.md Brand Personality. For multi-variant runs the register lives in per-variant `DESIGN-<id>.json`. ## 0.10.0 — `/stardust:uplift <URL>` one-shot presales orchestrator - NEW skill: `uplift` — produces three differentiated variants (A faithful + improvements, B "what if we amplified `<captured trait>`?", C "what if motion was part of the identity?" — fully cinematic) without further user coordination. - Closed catalog of 8 "what if…" captured-trait amplification candidates with disqualification clauses. - Load-bearing artifacts: `uplift-improvements.md` (5 specific weaknesses) and `uplift-questions.md` (catalog walk + picks + disqualifications). - Delegates to extract → direct → prototype with no new flags on the underlying skills. ## Plugin metadata - `tile.json` bumped 0.8.0 → 0.10.0; registers `uplift` + `prepare-migration` (the latter a pre-existing gap closed). - `.claude-plugin/plugin.json` bumped 0.9.0 → 0.10.0. ## Backward compatibility All existing sub-commands unchanged. Motion is opt-in via `--cinematic`. `uplift` is a new entry point with no parallel render path — it delegates to existing skills. Existing project trees gain new artifacts only when the new flags / skills are engaged. 🤖 Generated with [Claude Code](https://claude.com/claude-code) |
||
|
Paolo Mottadelli
|
1ab6833f1e |
fix(stardust): make migrate output genuinely zip-and-deploy portable (#133)
* fix(stardust): make migrate output genuinely zip-and-deploy portable The 0.7.0 contract claimed "self-contained, zip-and-deploy" but emitted root-absolute paths (/assets/..., /beers/) plus directory-only nav links. Those work on a webserver at the host root only — they 404 on file:// and at any subpath. This PR delivers what the claim promised: - Output path is URL-literal: /beers/ -> beers/index.html, /about-us/history.html -> about-us/history.html. No more synthesizing /about/index.html from a slug. - Page map (state.json.migrate.pageMap[]) built once per run; per-page rendering and internal-link rewriting both consume it. - Reference shape is depth-aware relative: prefix is ./ for depth-0 pages, ../ * depth otherwise. Works on file://, at host root, and at any subpath without rewriting. - Phase 3 broadens the self-containment check into a full portability audit: refuses absolute internal refs, refuses directory-only nav, runs the new pageMap + file:// fixtures. - New fixtures: file-protocol-audit.mjs (Playwright file:// round-trip) and pagemap-audit.mjs (static pageMap consistency). - asset-bundling.md and migrate-output-format.md aligned to the new depth-aware-relative shape; state.json contract gains pageMap[]. Plugin bumped 0.7.0 -> 0.7.1. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * review: fix four flagged correctness issues - Directory-only-nav regex now catches `./beers/` (the cited example). Old pattern required a-z as first char; `./beers/` starts with `.` and would have slipped through. New pattern accepts ./, ../, /, or a bare segment, and excludes scheme-bearing URLs (no false positive on https://google.com/). - Add a second grep for absolute url() references inside inline style attrs and <style> blocks (and CSS files); the original href/src grep missed these. - SKILL.md Outputs row no longer claims the sidecar lives at <source-url-path>/_meta.json (nonsense for .html-leaf pages). Replaced with a row that points at the canonical sidecar rule. - migration-procedure.md § Page map now explicitly says same-origin absolute URLs are stripped to their path portion before pageMap lookup (per content-preservation.md § Internal link rewriting), closing the loophole that would let <a href="https://origin/foo/"> pass through verbatim. Regex patterns verified against a test fixture before commit. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * review: update migrate-self-contained-bundle eval to 0.7.1 contract The eval was pinned to 0.7.0's root-relative /assets/<subpath> shape. This PR's contract change to depth-aware relative paths would have made the eval fail against correct 0.7.1 output. - detection_src_href criterion now asserts <prefix>assets/<subpath> (not /assets/<subpath>). - Renamed prefix_resolution_root_relative -> prefix_normalises_authoring; description flipped from "reference is left as-is" to "authoring is normalised to the depth-aware emit form". - Added 4 new criteria: depth_aware_paths (6pts), pagemap_constructed (6pts), internal_links_via_pagemap (5pts), portability_audit_passes (13pts). - Setup uses real source URLs that exercise the URL-literal rule: /, /beers/, /about/history.html (preserve .html leaf), /docs/api/ (depth-2 nested). - Prototype HTML now carries internal nav in three authoring styles (root-absolute, same-origin absolute, directory-only) plus a broken-link case, exercising the page map rewrite. - Run 1 expected behavior rewritten to expect depth-aware emission ('./assets/...'), pageMap built before render, broken-link flagging, and the 6-step portability audit passing. - Run 4 (cross-page dedup) updated so the second page emits its shared asset with its own '../' prefix while the underlying file on disk is a single copy. - Weights re-balanced to sum to 100; portability_audit_passes gets the largest weight (13pts) since it backs the contract claim. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
27dd06341c |
feat(stardust): migrate produces self-contained zip-and-deploy bundle (#132)
* feat(stardust): migrate produces self-contained zip-and-deploy bundle Bundles every referenced asset into stardust/migrated/assets/ and rewrites HTML/CSS references to root-relative /assets/<subpath> so the migrated tree is fully self-contained. Adds asset-bundling reference (six detection shapes, nine edge cases), output-format contract with state.json.migrate block (selfContained: true), --clean and --pin-timestamp flags, and an eval covering the six acceptance criteria. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * review: fix count, --clean correctness, document asset-missing rename - Bump totalAssetsBundled 6 → 7 in eval task.md (matched bundledAssets[] length). - --clean now implies --force so the bundledAssets Set is the complete union across every page; idempotent-skipped pages' assets are no longer at risk of being deleted as stale. Updated SKILL.md flag description, asset-bundling.md § Stale asset cleanup, the Idempotent-skip interaction note, and the eval's run-3 expected behaviour + criterion. - Document the contentDeviations / migrationDecisions kind rename from "media-missing" to "asset-missing" in migrate-output-format.md § Backwards compatibility — consumers should accept either string until they migrate. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * chore(stardust): bump plugin version 0.6.0 → 0.7.0 Self-contained zip-and-deploy bundle is a minor-version addition: new --clean and --pin-timestamp flags, new state.json migrate block (forward-compat), new asset-bundling reference contract, provenance kind rename (media-missing → asset-missing). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * chore(stardust): bump palette library version v0.6.0 → v0.7.0 Keep the palette library version aligned with the plugin version. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
bfe537dbdd |
feat(stardust): ia-fidelity axis + 10 non-slop disciplines + approval fold-back (#131)
Coordinated three-part change: Part I adds a strategic ia-fidelity axis pinned at direct time (verbatim → A1/A2/A3 surface forks vs reimagined → A+B+C role-differentiated). Part II adds 10 disciplines at three phase boundaries (brief-time 1-5, craft-time 6-8, quality gates 9 + convergence detector 10) with the four wasatch dry-run friction carve-outs folded in. Part III lands Phase 5 approval fold-back (site-wide / page-local / don't-fold) with verbatim no-op gating. Also sweeps remaining Phase 5.5 → 2.7 references across the plugin docs. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
5d47099881 |
feat(stardust)!: sweep — drop doctor, fossils, opt-out flags, live integration
* feat(stardust)!: sweep — drop doctor.md, F-NNN fossils, opt-out flags, live integration
Four related cleanups bundled. Each independently small; together
they shrink the docs by ~600 lines and remove three doc-rot
patterns at once.
1. Delete doctor.md (493-line spec for an unimplemented script).
The Festool-incident concerns it was scoped to are now handled
by concrete write-time contracts (validateProvenance, content-
sourcing hierarchy, anti-toolbox audit). No external references.
2. Drop F-NNN feedback fossils + version-changelog paragraphs.
STARDUST-FEEDBACK.md doesn't exist in the plugin; the F-001 /
F-002 / F-009 / F-015 / F-016 / F-018 citations were opaque
tribal vocabulary. Kept the rules they cited; dropped the
IDs. divergence-toolkit.md changelog paragraphs (v0.1 → v0.2,
v0.4 → v0.5, v0.5 → v0.5.1) collapsed.
3. Slim flag surfaces. Every --no-* / --skip-* / --allow-* /
--refresh-stale flag dropped from prototype and migrate; the
underlying quality gates are now mandatory. Manual chat
overrides ("ship as-is", "accept the P1 findings") remain
available and are recorded in _provenance verbatim.
prototype flags: 10 → 5 (<slug>, --all, --prep, --canon-from,
--publish-sample).
migrate flags: 8 → 5 (<slug>, --all, --force, --require-approved,
--strict-canon).
4. Drop the impeccable:live integration. Live was documented as
the "default" iteration path but wasn't used — chat-driven
command invocation is the actual default. Phase 4 collapses
from 6 steps to 3 (open file → mark prototyped → report).
Iteration paths drop from 3 to 2 (chat-driven, direct
invocation). Live remains usable as an external tool the user
invokes directly against the proposed file; stardust no
longer pretends to drive its poll loop.
BREAKING CHANGE: prototype no longer accepts --no-iterate,
--no-critique, --skip-adapt-audit, --adapt-variant, --no-hamburger,
--refresh-stale, or --no-impeccable. migrate no longer accepts
--refresh-stale, --allow-placeholder, or --skip-adapt-audit.
The mobile-adapt audit, content-sourcing scan, placeholder gate,
and critique + audit pass are now mandatory; manual chat overrides
replace the flag-based bypasses.
The live driver loop (Phase 4 steps 4-6: config.json setup, event
streaming, carbonize cleanup, iteratedVia: impeccable:live
provenance) is gone. The proposed file's iteratedVia provenance
now records the named impeccable commands run against the file,
not a live sessionId.
stardust bumped to 0.6.0.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* chore(stardust): sync tile.json version to plugin.json (0.6.0)
The tile manifest's version was stuck at 0.1.0 while the plugin
moved to 0.6.0. Keep them in sync so tessl-side catalog metadata
matches the marketplace.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
|
||
|
Paolo Mottadelli
|
33dd9037f3 |
stardust: mobile-nav collapse + drop before/after viewer (#127)
* stardust: mobile-nav collapse pattern + drop before/after viewer
Two related changes to stardust:prototype landing together.
Mobile-nav collapse (additive):
- New stock CSS-only hamburger pattern auto-applied by Phase 5.5
when the Mobile-adapt audit refuses at 360px (overflow,
sub-11px nav font, or sub-10px nav gap).
- New audit conditions documented in skills/prototype/SKILL.md
§ Mobile-adapt audit; checked at runtime via the Playwright
snippet at skills/prototype/fixtures/mobile-nav-audit.mjs.
- Stock pattern doc at skills/prototype/reference/mobile-nav-collapse.md
carries the copy-pasteable HTML+CSS+JS, the ARIA checklist,
the post-injection source-order check, --no-hamburger opt-out,
and the alternative-pattern vocabulary.
- New data-nav-collapse="hamburger" attribute (v2.2 additive) on
<header> when the pattern is applied.
- Positive + negative fixtures verify the audit end-to-end.
Drop the per-page before/after viewer:
- Phase 3 (Compose the viewer) removed; the proposed file is
opened directly in the browser.
- Approval is now chat-only: "approve <slug>" or "approve".
- skills/prototype/reference/before-after-shell.md renamed to
proposed-file-shell.md and stripped to the proposed-file
contract only (viewer half, iframe resolution chain, action
buttons, and stash mechanic all dropped).
- state-machine.md prototypePath repointed at -proposed.html;
doctor.md D-104 pairs -proposed.html with -shape.md;
artifact-map.md drops the <slug>.html row; publish-sample.md
simplifies variant discovery (no more "is not the viewer"
disambiguation); evals/prototype-before-after rewritten to
expect a single proposed file.
BREAKING CHANGE: stardust:prototype no longer composes the
before/after viewer at stardust/prototypes/<slug>.html.
Anyone scripting on top of that file existing on disk should
switch to stardust/prototypes/<slug>-proposed.html (the
prototypePath in state.json now points there directly). The
approval gesture moves from the viewer's Approve button to a
chat-only "approve <slug>" / "approve".
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* stardust: bump version to 0.5.0
Minor bump for the breaking change in
|
||
|
Paolo Mottadelli
|
2da4f830ad |
stardust: fix dependencies schema to use name@marketplace string (#126)
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
90be5b7bf2 |
stardust: fix cross-marketplace dep reference to impeccable
The `marketplace` field in plugin dependencies and the `allowCrossMarketplaceDependenciesOn` allowlist must use the marketplace's `name` (`impeccable`), not its GitHub repo path (`pbakaus/impeccable`). The mismatch caused Claude Code to flag the stardust plugin manifest as invalid on install. Bumped to 0.4.1. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
69068cee1c |
stardust: bump version to 0.4.0
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
48ec3970ef |
stardust: close open clusters + declare cross-marketplace impeccable dep (#120)
* feat(stardust): close open clusters — publish-sample URL, mobile-adapt phase, critique+audit, add-variant Four items from the 2026-05-05 review session covering UPS F7+F9, NVIDIA F8, and dogfood-feedback 2026-05-03-a / JFK F14. All four shipped together because they overlap in the same SKILL files. - publish-sample showcase URL fix (UPS F9): the printed URL pattern pointed at a route that doesn't exist (samples/<slug>/), so every published sample carried a broken link in the PR body and final report. Updated to the actual single-page-viewer route (samples/brand.html?slug=<slug>#<variantId>) and synced the upstream default to paolomoz/stardust-site. - Mobile-adapt as a pipeline phase (dogfood-feedback 2026-05-03-a + JFK F14): new prototype Phase 5.5 runs $impeccable adapt against the approved variant on every approval; mobile-adapt audit at prototype, migrate, and publish-sample refuses files with no <meta viewport>, zero @media (max-width: ...) rules, or no mobile-targeted breakpoint at <= 640px. --skip-adapt-audit escape hatch threads through all three. Closes the failure mode where the lovesac.com showcase shipped variant B to mobile visitors with bracket motif crowding, overflowing trust band, and no hamburger. - Phase 2.5 critique + audit (NVIDIA F8): renamed from "Validate via critique" to "Validate via critique + audit"; runs both impeccable validators in parallel and gates prototyped status on P0/P1 from either. Added brand-faithful inversion auto-dismiss against DESIGN.json#extensions.divergence.brand_faithful_inversions[] so Mode A renders don't drown the report in known false positives (Arial fallback flagged as overused-font, eyebrow uppercase as all-caps body, etc.). publish-sample's submitter checklist updated to cover both validators. - direct --add-variant <name> (UPS F7): incremental variant addition against the existing direction without re-running Phase 1 reasoning or stale-flagging existing prototypes. Procedure documented; per- field inheritance rules table specifies what's pinned (palette, typography, IA-priorities), what extends (componentStyle, divergence inversions), and what's variant-local (narrative, voice DOs, the amplified-trait anchor). Fixes the workflow where variant A ships, user later asks for B and C, and --re-direct is too heavy. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * chore(stardust): declare cross-marketplace dependency on impeccable Whitelist pbakaus/impeccable in marketplace.json and declare it as a dependency in stardust's plugin.json so stardust can formally depend on the impeccable plugin hosted in a different marketplace. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Robert Munteanu
|
eb78c8a148 |
fix(stardust): fix tile.json to pass tessl-lint (#116)
- Add missing version/summary fields and convert skill paths from strings to objects to match the required tile.json schema. - tessl-lint rejects colons in skill names; only lowercase letters, numbers, and hyphens are allowed. - Skill keys must match the name field in each SKILL.md; drop the stardust- prefix since the SKILL.md files declare short names. --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
5bb2150e37 |
feat(stardust): close open clusters — publish-sample URL, mobile-adapt phase, critique+audit, add-variant (#113)
Four items from the 2026-05-05 review session covering UPS F7+F9, NVIDIA F8, and dogfood-feedback 2026-05-03-a / JFK F14. All four shipped together because they overlap in the same SKILL files. - publish-sample showcase URL fix (UPS F9): the printed URL pattern pointed at a route that doesn't exist (samples/<slug>/), so every published sample carried a broken link in the PR body and final report. Updated to the actual single-page-viewer route (samples/brand.html?slug=<slug>#<variantId>) and synced the upstream default to paolomoz/stardust-site. - Mobile-adapt as a pipeline phase (dogfood-feedback 2026-05-03-a + JFK F14): new prototype Phase 5.5 runs $impeccable adapt against the approved variant on every approval; mobile-adapt audit at prototype, migrate, and publish-sample refuses files with no <meta viewport>, zero @media (max-width: ...) rules, or no mobile-targeted breakpoint at <= 640px. --skip-adapt-audit escape hatch threads through all three. Closes the failure mode where the lovesac.com showcase shipped variant B to mobile visitors with bracket motif crowding, overflowing trust band, and no hamburger. - Phase 2.5 critique + audit (NVIDIA F8): renamed from "Validate via critique" to "Validate via critique + audit"; runs both impeccable validators in parallel and gates prototyped status on P0/P1 from either. Added brand-faithful inversion auto-dismiss against DESIGN.json#extensions.divergence.brand_faithful_inversions[] so Mode A renders don't drown the report in known false positives (Arial fallback flagged as overused-font, eyebrow uppercase as all-caps body, etc.). publish-sample's submitter checklist updated to cover both validators. - direct --add-variant <name> (UPS F7): incremental variant addition against the existing direction without re-running Phase 1 reasoning or stale-flagging existing prototypes. Procedure documented; per- field inheritance rules table specifies what's pinned (palette, typography, IA-priorities), what extends (componentStyle, divergence inversions), and what's variant-local (narrative, voice DOs, the amplified-trait anchor). Fixes the workflow where variant A ships, user later asks for B and C, and --re-direct is too heavy. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
ed46712dde |
stardust: extract integrity & capture-fidelity hardening from 2026-04-29 → 2026-05-04 dogfood (#112)
* feat(stardust): provenance integrity — synthesis guard, cascade validation, full body capture Closes a class of silent corruption where extract --prep silently synthesized page records and the cascade ran four phases on fabricated data before being caught by a meta-question (2026-04-30 lovesac.com). - extract refuses to mark a page extracted without renderedBy:"playwright" + fetchedAt + waitMs>0 + waitMode + httpStatus; new ProvenanceMissing error class; sub-agent prompts must forbid synthesis by name and return a per-page evidence table. - validateProvenance(page) specced in state-machine.md; called at setup of direct --prep, prototype, migrate; cascade guard runs between prepare-migration Phase 1 and Phase 2 with explicit abort message. - innerText captured in full — no truncation. Reference scripts must not slice (the v0.2 .slice(0, 4000) silently dropped most body content on long-form pages and broke module detection downstream). - Schema additions for sections[].body, lists, qa, quotes so migrate can render real body copy under each heading instead of falling back to placeholder-with-signature throughout. - Mandatory per-page evidence table in extract Phase 6 with live: yes/no column; Provenance: N/N live line in prep summary so the contract is visible at a glance. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(stardust): extract reliability — Akamai headed fallback + consent pre-flight Two unrelated context-setup gaps surfaced in 2026-05-03/04 dogfood that silently produced bad captures. - Bot-management fallback (UPS, ups.com/gb): Akamai rejects bundled- chromium-default headless on its TLS/H2 fingerprint with ERR_HTTP2_PROTOCOL_ERROR, before any JS runs. Recipe now specifies headless: false, channel: 'chrome' as the retry path; playwright- recipe.md gains a Bot-management fallback section, extract setup switches modes on the protocol error without retrying headless, and the in-page fetch pattern is documented for sub-resource fetches that need to inherit the JA3 fingerprint. - Consent banner pre-flight (NVIDIA, nvidia.com): every page's screenshot was overlaid by the OneTrust banner and the voice table was dominated by 'Manage Settings' / 'Reject All' / 'Accept All' once-per-page. Recipe now specifies an API-first dismissal (OneTrust / Cookiebot / Didomi / Osano) with selector fallbacks, --no-consent-dismiss opt-out, and consent.method recorded in _crawl-log.json. Side-effect caveat documented: the dismissal can activate other gated scripts (analytics, geo-IP, A/B) and the result may differ from a no-dismiss capture. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(stardust): extract capture fidelity — fonts, icon font, CSS-bg pseudo, hero image, logo rect-filter Five capture-fidelity gaps surfaced across the 2026-04-29 → 2026-05-04 dogfood window. Each silently produced a thinner or wrong brand surface that downstream prototypes then rendered against. - Logo locator step 2 (UPS): rect filter (height ≥ 32, width ≥ 40, top ≤ 200) + aspect filter [0.5, 3.0] + rank by proximity to 1.5 ideal. Stops promotional icons that share the brand-name substring from being picked as the masthead logo. - CSS-background hero (UPS): pseudo-element walk on ::before / ::after generated content, which document.querySelectorAll('*') doesn't reach. Performance-bounded by the prior visible-size filter. - voice.heroImage elevation (UPS): single canonical hero per home, resolved by viewport-top + visible-area + aspect-ratio filters with explicit DOM > CSS-bg > pseudo tie-breakers. Ends downstream re-derivation from a 16-image list. - Font file capture via network-intercept (JFK): every woff2/ttf/otf saved to assets/fonts/, recorded in _brand-extraction.json#type.files[] with @font-face metadata + licensing flag. Closes the gap where Mode A prototypes silently fall back to system-ui on private-cut brands. - Icon-font detection (JFK): query [class^="icon-"] + ::before font-family + codepoint, build iconClass → codepoint table, save the font file. Closes the emoji-fallback render path on Chakra/Material/government sites. - Module-detection signal priority (sliccy): explicit ranking — headings → CTAs → cssBackgrounds → forms → componentsByLandmark → innerText substring. Substring search drops to corroboration only; was the source of the 2026-04-29 under-detection. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
9cfed1145a |
feat(stardust): replace v1 with v0.3.0 (redesign on top of impeccable) (#111)
v1 (four-stage greenfield: brand → briefings → wireframes → prototype) is preserved at tag stardust--v0.1.0. v0.3.0 is a complete refactor with a new surface (extract → direct → prototype → migrate) and a hard runtime dependency on impeccable. - plugins/stardust: replace contents with v0.3.0 source from paolomoz/stardust-2 (skills, evals, README, tile.json, plugin.json) - plugins/stardust/.claude-plugin/plugin.json: bump to 0.3.0; add repository and v2-relevant keywords - plugins/stardust/skills/*/SKILL.md: add license: Apache-2.0 to frontmatter - .claude-plugin/marketplace.json: update stardust entry with new description, version, category, keywords, author, homepage, repository, license (mirrors adobe-for-creativity entry shape) - plugins/stardust/README.md: status bumped to v0.3.0; v1 link repointed to the stardust--v0.1.0 tag Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
91589bfb0d |
docs(stardust): refresh README with identity visuals and voice
Adds cover, pipeline, and logo SVGs under docs/ and rewrites the README to lead with the identity (brief + seed = star) while keeping all instructional sections. Adds a short "how it wants to be spoken about" section so docs and tooling on top of stardust can match register. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
Paolo Mottadelli
|
a183041f6e |
feat(stardust): publish design-phase skills plugin (#92)
* feat(aem-design): scaffold plugin skeleton - plugin.json (Apache-2.0, Adobe author, v0.1.0) - tile.json listing the 5 design skills - empty skills/ and evals/ (populated by subsequent commits) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design): import hardened skills from stardust-plugin 5 design skills (aem-design navigator, brand, briefings, wireframes, prototype) plus _shared/ (preflight, skill-contract, soft-deps, fallback-brainstorm, fallback-critique). All skills use /aem-design:* slash commands, aem-design:provenance markers, no absolute paths, and no EDS dev-server references. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * test(aem-design): import eval packs 5 packs (aem-design-navigator, brand-extract-from-url, briefings-from-prompt, wireframes-render, prototype-iterate). Each has task.md (setup, prompt, expected behavior) and criteria.json (weighted rubric summing to 100). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * docs(aem-design): plugin README Install, skill table, soft-dep notes (superpowers, impeccable), AGENTS.md snippet for activation reinforcement, and out-of-scope clarification (EDS build phase is separate). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(marketplace): register aem-design Insert alongside the other AEM entries, immediately before aem-edge-delivery-services to group the design-phase plugin with its downstream EDS implementation companion. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * docs(readme): add aem-design section and install command Adds the Designing-with-aem-design subsection with the 5 skills (navigator + brand, briefings, wireframes, prototype) under the AEM section, placed before Developing with Edge Delivery Services to reflect pipeline order. Adds /plugin install aem-design to the top install block. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * docs(aem-design): add "When to use" / "Do NOT use" blocks per Tessl rubric Each SKILL.md now names the triggers that should invoke it and defers to sibling skills for out-of-scope requests, so agents with weaker description matching hand off cleanly instead of overreaching. * fix(aem-design/brand): enforce _provenance contract and extract real logo to aem-design/assets/ Follow-ups #1, #3, #4 from docs/superpowers/aem-design-followups.md. - brand-profile.json now requires `_provenance` as the first top-level key with generated_by, source, extraction_method, synthesized_inputs (enumerating LLM-filled fields) and screenshots. Legacy `extraction` block removed. Aligns with _shared/skill-contract.md. - brand skill now extracts the real logo on a deterministic fallback path: inline <svg> in header/nav → <img> with logo class/src → apple-touch-icon / og:image → favicon → synthesized placeholder (last resort). Output always lands at aem-design/assets/logo.<ext> — never icons/ — so the design phase stays EDS-agnostic. brand-profile.json's logo object now records path/format/source. brand-board.html references the asset file via <img>, not inline SVG. - brand-extract-from-url eval: provenance_stamped criterion tightened to assert the new contract shape; new real_logo_asset criterion (10 pts) asserts the file exists under assets/ with logo.source set; weights rebalanced to sum to 100. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(aem-design): honor soft-deps delegation across skills (briefings, brand, wireframes) Explicit decision tree at each soft-dep boundary: detect the peer slash command, delegate when present, otherwise announce the fallback once per session and run the inline pattern. Silent inline interviews are now a bug. - briefings/SKILL.md Phase 2 Step 1: rewrite discovery to branch on /brainstorm registration. Either delegate with visible hand-off or announce fallback once. - brand/SKILL.md conversational path: same pattern for brand discovery. - wireframes/SKILL.md Phase 1 Step 4: same pattern for multi-page IA planning via /write-plan. - evals/briefings-from-prompt: rename fallback_announced_once -> soft_deps_honored with tightened description covering both delegation and announcement paths. Follow-up #5 from PR #92 smoke test. * feat(aem-design): briefings default to structured shape with [TBD] placeholders Previously the skill asked the user to pick a fidelity level (prompt-only / structured / full) before drafting. This created an unnecessary gate — the user had to front-load a planning decision before they had any output to react to. Now every briefing emits the same structured shape (Intent + Audience + Key Messages + CTAs + Tone), with `[TBD]` written verbatim in sections the user hasn't committed to. Downstream skills synthesize on-brand content for `[TBD]` fields and stamp provenance. Fidelity becomes a consequence of how much the user fills in, not a tier they select upfront. - briefings/SKILL.md Phase 1: drop the fidelity question, default to structured, document [TBD] semantics. Phase 2: draft in structured shape always. - briefing-format.md: replace three-tier fidelity table with structured+full shape table. Remove `fidelity` frontmatter field. Update Rules to require all five structured sections (with [TBD] valid as section body). - evals/briefings-from-prompt: add `no_fidelity_gate` criterion (10), retune weights to keep total=100. Tighten briefing_sections and provenance_stamped descriptions. Rewrite task.md expected behavior. Follow-up #6 from PR #92 smoke test. * feat(aem-design): prototype skill offers 1-4 variants to choose from Early variant choice is cheaper than late iteration — committing silently to one design direction means that changing direction after the user has asked for type-scale tweaks throws away the tweaks. The soft-gate philosophy already embraces "synthesize and stamp"; variants are the logical extension. - prototype/SKILL.md Phase 0: ask "How many variants? (1-4, default 1)". N=1 preserves current behavior. N>=2 renders {page}-a.html, {page}-b.html, ... each exploring a load-bearing axis: type-voice, density, color-energy, imagery-role. Each variant's provenance block carries a `variant_direction` line. - Phase 3 (Serve): for multi-variant, present a comparison table (direction / key visual move / risk) and ask the user which to take forward. - Phase 4 (Iterate): iterate only on the chosen file. Unchosen variants stay on disk as design-history reference. - Artifacts table documents both {page}.html and {page}-{letter}.html shapes. - evals/prototype-iterate: add Scenario B (3 variants + choose B + iterate). New criteria: variant_count_asked (5), variant_directions_distinct (8), iteration_targets_chosen_variant (8). Retune existing weights to keep total=100. task.md extended with Scenario B expected behavior. Follow-up #7 from PR #92 smoke test (minimum ambition per plan). * feat(aem-design/brand): add divergence toolkit and make reference-required the default Push back against LLM-driven visual convergence across brand generations by establishing three within-session mechanisms: 1. New _shared/divergence-toolkit.md (v0.1, self-audited) houses: - Default-moves budget (max 3 per profile) listing recurring LLM defaults across typography, motifs, voice rules, structure, and palette roles - Seed lists (decade x craft tradition x cultural register) used as hard constraints when no reference is provided; deterministic-random from MD5(brand-name + date) for audit reproducibility - Named font decks (editorial-archival, retro-italian, zine-maximalist, swiss-modernist, etc.) replacing ad-hoc font picks - Role-naming rule forbidding generic palette role slots - Reference-use discipline for listicle inputs (extract all trends, tag in/off-toolbox, require off-toolbox per variant) 2. brand/SKILL.md Phase 1: reference-required by default. Conversational input drops to option 4 (explicit escape hatch) with a verbatim warning citing the toolkit. When accepted, rolls a seed and stamps _divergence.divergence_warning=true. 3. brand/SKILL.md Phase 2: .impeccable.md authorship stamping. Every file opens with frontmatter declaring authored_by (designer|synthesized) and strength. Adds a "last chance" interview prompt before synthesis so downstream skills don't enforce LLM defaults as designer taste. Schema additions in brand-profile-schema.md: - New optional _divergence block as the second top-level key (captures seed, chassis, font_deck, anti_toolbox_count/hits, justifications, off_toolbox_moves, references_used with listicle tagging, warning flag) - Role-naming enforcement on new writes; old profiles remain valid reads - Optional sibling "use" field on color entries for technical qualifiers so role stays brand-native Backward-compatible: missing _divergence block is valid; existing profiles with generic role names are still read. No downstream skill is required to hard-fail on these additions. Addresses convergence observed when the same LLM regenerates brand boards for unrelated categories (e.g., horror publisher vs. Italian pantry brand sharing stencil typography, hazard stripes, rotated stamps, and identical palette role taxonomies). See tmp/e2e-3 divergence report for analysis. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/brand): divergence toolkit v0.2 (I1+I2+I3+I5+I7) Bundle five within-file edits to _shared/divergence-toolkit.md after a 4-run test (e2e-4/5/6/7) surfaced specific gaps in v0.1. Changes: I1 - Expanded § 1 Default-Moves List: - Added Triplet-cadence copy to Voice-rule moves (detected in all 4 test runs: "Same press. Same shop. Same eight years." / "Dated, classified, cross-linked." / "Saturday. Bar Beach. We dance.") - Added Collage-maximalism kit to Motif moves (the assistant's own archetype, reached for in fresh sessions without context) - Added Generic-2026-SaaS silhouette and Stat-callout bar to Structural moves (Linear/Notion/Stripe aesthetic; observed in Orrery run) - New Palette-family subsection covering the "archival editorial" palette that recurred in 3 of 4 test runs (cream ground + warm-family saturated accent + muted earth-tone secondary), cream-as-default page ground, brutalist-pomodoro, and dark-mode-editorial I2 - Replaced the 3-free-hit budget with per-hit justification: - Every hit from § 1 now requires a brand-specific justification that would not transfer unchanged to an arbitrary other brand - Retired the budget because the 4-run test (Yadda Dey, 4 hits) showed the LLM comfortably fills it with defaults and rationalises excess I3 - Moved sensible universal voice rules out of § 1 into new § 7: - Banned marketing adjectives list (artisanal/crafted/premium/curated/ beloved/cozy/warm/inviting/thoughtfully/delightfully/lovingly/ uncompromising/bespoke) - No exclamation points except italic quoted outbursts - Em-dashes welcome / semicolons earned These are opt-in house standards, copied into voice.rules[] when the brand or seed implies them, and do NOT count against the budget I5 - Added new § 2.5 Dimension Weighting: - decade drives type deck + cultural references + display register - craft drives texture + motif idiom + print artifacts - register drives voice stance + structural metaphor + IA - Within-run variants use dimension dominance (A decade-lead, B craft- lead, C register-lead) to produce meaningful variant divergence - Added "Ground color by seed" table: cream belongs only to specific seeds (1920s-1960s x letterpress/riso/field-guide/travel-brochure); 1990s x legal-contract belongs on pale gray; 1970s x enamel-sign belongs on saturated ground. Cream as default is now a § 1 hit. I7 - Appended self-audit rationalisation check to Enforcement: - Hit audit: "would a reviewer agree this is a hit?" (A 'travel decal' that is a rotated circular stamp IS a rotated circular stamp) - Off-toolbox audit: is it a genuine invention or a dressed-up default? - Triplet-copy audit: max one triplet per page - Adjustments recorded in _divergence.audit_adjustments[] Status banner updated to v0.2 - self-audited - run-tested with a v0.1 -> v0.2 changelog. Backward compatibility: _divergence.toolkit_version carries the version so downstream skills can tell which list was in effect. Addresses gaps surfaced in: - /Users/paolo/excat/tmp/e2e-3/aem-design-divergence-test-assessment.md - /Users/paolo/excat/tmp/e2e-3/aem-design-toolkit-v0.2-improvements.md Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/brand): tighten role-naming + remove cream default (I4+I8) Close two palette-vocabulary loopholes observed in the 4-run test. Both edits are small and compound with the divergence toolkit v0.2 shipped in the previous commit. I4 - brand-profile-schema.md Role Naming: token-level match The v0.1 rule forbade specific values as the sole role name (Primary, Secondary, Alarm, Warning, Shadow, Hardware, Accent, Background). That string-equality check missed compound smuggling like "Accent Gold" / "Primary Red" / "Brand Blue" — the generic taxonomy sneaks in alongside a color name and passes validation. v0.2 upgrades to token-level whole-word case-insensitive match. Any role value containing a token from the forbidden list is rejected, including compounds. New tokens added to the list: Neutral, Brand, House. - Forbidden on new writes: Primary, Secondary, Tertiary, Alarm, Warning, Danger, Shadow, Hardware, Ink, Accent, Background, Neutral, Brand, House (anywhere in the role value, not just as the whole value) - Technical intent continues to live in the sibling 'use' field, which is allowed to contain the forbidden tokens - The color's 'name' field is also unrestricted - a color called "Shadow of Pomodoro" is fine; the 'role' field is the one that must be brand- native Existing profiles with generic role names remain valid reads. The rule applies only to new writes. I8 - brand-board-template.md: remove cream default substrate The v0.1 template said: "Light background (#faf9f6 or similar warm neutral) for the page body". That instruction drove cream backgrounds in 3 of 4 test runs regardless of whether the brand's seed called for cream. Shita-gawa's seed was 1990s x Letterpress x Legal contract (pale gray substrate per toolkit v0.2 section 2.5) and still landed on cream because the template told it to. v0.2 replaces the instruction with a brand-derived-ground rule that points at the divergence toolkit's Palette-family moves (section 1) and Ground color by seed table (section 2.5). Example seeds that want cream are named inline; example seeds that should NOT default to cream are also named. The dark-sections rule is softened: inversion happens only when the brand's palette supports an ink-or-deeper-than-ground surface, not as a blanket template default. Addresses gaps identified in: - /Users/paolo/excat/tmp/e2e-3/aem-design-divergence-test-assessment.md - /Users/paolo/excat/tmp/e2e-3/aem-design-toolkit-v0.2-improvements.md (I4, I8) Canonical test after both commits: a Shita-gawa re-run should either produce a non-cream ground, or stamp _divergence.anti_toolbox_hits[] with a "Cream/paper as default page ground" entry carrying a brand-specific justification. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/brand): chassis library (I6) Split brand-board-template.md into a data contract and a chassis library, ship 4 chassis specs, and update brand/SKILL.md Phase 3 to pick a chassis per run. This is the single-largest remaining convergence axis: v0.2 boards diverged in palette/type/motif but shared one skeleton. Chassis addresses that. New files (brand/reference/chassis/): classic-archive.md - v0.1 default preserved as one option; numbered sections, sticky top nav, masthead with metadata stamp, horizontal grid. Appropriate for archival/institutional/editorial brands. Reference: Pentagram annual report, MacGuffin. dashboard.md - fixed left sidebar, cards-as-widgets, meters / gauges / status chips. Appropriate for operational/tool brands and seeds with railway-timetable, broadcast-captioning, sports-scorecard, repair-manual, hospital-discharge registers. Reference: Linear admin, Bloomberg Terminal overview. magazine.md - masthead with Vol/Issue/date, column text, pull quotes, drop caps, magazine department heads (FEATURE / ESSAY / LETTER). Appropriate for tabloid, memoir, auction-catalogue, travel-brochure registers. Reference: Apartamento, The Gentlewoman. pinboard.md - no sticky nav, cluster-based spatial arrangement, handwritten labels, scrapbook overlap. Appropriate for zine register and collective/community/youth brands. Reference: La Palatine, Cy Twombly workbook. Carries explicit caveat that picking pinboard does NOT excuse hitting the Collage-maximalism kit from divergence-toolkit.md section 1 Motif moves. Each chassis spec covers: intent, navigation, section order, hero shape, eyebrow conventions, spacing rhythm, typography register, motif placement, what it is NOT, and example references. Structural, not taste-prose. Modified files: brand/reference/brand-board-template.md - rewritten as a data contract (what must be rendered) plus a chassis picker section. Adds explicit chassis-selection rules with seed-informed preference table. Previous "Required Sections" block now a plain bullet data-contract list. brand/SKILL.md Phase 3 - rewritten to pick a chassis before rendering: 1) designer override, 2) byte[3] hash from seed.hash_input mod chassis count, 3) seed-preference override, 4) fallback to classic-archive. Records _divergence.chassis and _divergence.chassis_pick_reason. brand/reference/brand-profile-schema.md - adds chassis_pick_reason field to the _divergence block. chassis field description expanded to name the four available chassis. After this commit, chassis selection is auditable: _divergence.chassis and _divergence.chassis_pick_reason tell a reviewer exactly which chassis landed and why. The expected outcome on a 4-run test (same brands as e2e-4/5/6/7) is at least 3 different chassis across the 4 boards — structural divergence, not just surface divergence. Addresses Improvement I6 in: - /Users/paolo/excat/tmp/e2e-3/aem-design-toolkit-v0.2-improvements.md Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * refactor(aem-design): chassis artifact-agnostic + prototype integration (v0.3) v0.2 shipped chassis as brand-board-only. The chassis files mixed presentation-pattern concerns (nav, hero pattern, spacing, type register) with brand-board-specific concerns (section list naming Palette/ Typography/Voice/etc). Prototypes still rendered on a single implicit chassis — the exact convergence problem v0.2 was supposed to address. v0.3 refactors chassis into a shared library both brand and prototype read, and adds a prototype chassis picker. Structural changes: 1. Move chassis files from brand/reference/chassis/ to _shared/chassis/. The library is artifact-agnostic; both brand and prototype skills now reference it. 2. Refactor each of the four chassis files (classic-archive/dashboard/magazine/pinboard): - Remove "Section order" subsection (brand-board-specific) - Rename "Hero shape" to "Hero pattern" (shape, not content) - Rename "Motif placement" to "Demo / visual presentation style" (applies to any demos, not just motifs) - Add explicit "this chassis describes a presentation pattern only" clarification at the top - Bump version 0.1 → 0.2, add scope: artifact-agnostic 3. brand-board-template.md: - Add "Section order per chassis (board-specific application)" block that maps the data contract to each chassis (this is where the brand-board-specific section order lives now) - Update chassis links from brand/reference/chassis/ to ../../_shared/chassis/ - Renumber the Template Generation steps to accommodate the added section-order-reading step 4. prototype/reference/design-guide.md: - Add introductory paragraph about the shared chassis library - Add new "Chassis selection" section: designer override via _divergence.prototype_chassis → inherit from brand's _divergence.chassis → fallback classic-archive - Add "Prototype section order per chassis" table with per-chassis landing-page section ordering (e.g. dashboard landing: hero cluster → feature grid → status strip → reservation card → FAQ tray) - Add cross-chassis rules: one primary CTA, briefing drives content selection, motifs/components applied within sections per chassis demo style, variants stay within the chosen chassis 5. prototype/SKILL.md Phase 2: - Add Step A "Pick the chassis" before the render step - Chassis recorded in the prototype's provenance block with chassis_pick_reason - Step B renamed, includes "Apply the chassis's presentation pattern" 6. brand-profile-schema.md: - chassis field description updated to note prototypes inherit by default and to point at the new _shared/chassis/ location - New optional prototype_chassis field for per-artifact override 7. brand/SKILL.md Phase 3: - Path references updated from reference/chassis/ to ../_shared/chassis/ - Clarified split: chassis spec is artifact-agnostic; section order lives in brand-board-template.md Result: a brand on magazine chassis produces a brand board AND a prototype both in magazine presentation pattern by default. Designer can override prototype_chassis when a given brand wants one chassis for internal documentation and another for public landing. Addresses the "chassis was brand-board-only" gap raised in review. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/brand): swap pinboard for broadcast-grid chassis (v0.4) Replace the pinboard chassis with broadcast-grid after review of v0.3's 4 chassis and 7 speculative mocks (see tmp/e2e-3/chassis-mocks/). Rationale: - Pinboard paired structurally with the Collage-maximalism kit listed in _shared/divergence-toolkit.md section 1. The chassis itself wasn't the problem (spatial cluster arrangement is a legitimate pattern), but in practice picking pinboard produced boards that then hit the anti- toolbox with rubber stamps, handwritten annotations, rotated date- blocks, and ripped edges. The chassis's very availability nudged the LLM toward its own "collage-maximalism" archetype — exactly the convergence the divergence work is trying to address. - Broadcast-grid fills a structural gap that pinboard didn't. v0.3's 4 chassis left no good home for brands whose seed register is railway-timetable, broadcast-captioning, sports-scorecard, liturgical- program, or pharmacy-insert — registers where the native form is a scoreboard / terminal / timetable. Dashboard was the fallback, but dashboard is cards-on-a-canvas; broadcast-grid is flat dense panes with a ticker. Different metaphor, different density. Changes: New: _shared/chassis/broadcast-grid.md (90 lines) - Intent: terminal / EPG / news-broadcast presentation - Navigation: top bar (channel + LIVE + tabs + clock) + ticker band - Hero pattern: multi-pane grid with 1.5fr hero pane + feed panes - Eyebrow conventions: pane title + source label, split by hairline - Spacing rhythm: 1px inter-pane rules, no radii, tight 14-16px pad - Typography: monospace-dominant, tabular numerals, condensed sans for pane titles; dark surface by default, light variant valid - Demo style: each demo rendered as a live-feed pane (table/list + source label), not as individual cards on a canvas - Explicit NOT list vs dashboard / classic-archive / magazine - Accessibility note on high-density contrast requirements Removed: _shared/chassis/pinboard.md Updated chassis selection and section orders: brand/reference/brand-board-template.md: - Chassis list now alphabetical: broadcast-grid, classic-archive, dashboard, magazine - Hash-based default mapping updated: 0 -> broadcast-grid, 1 -> classic-archive, 2 -> dashboard, 3 -> magazine - Seed-informed preference table rewritten: - railway-timetable / broadcast-captioning / sports-scorecard / liturgical-program / pharmacy-insert -> broadcast-grid OR dashboard - zine register moved to magazine (was pinboard in v0.3) - legal-contract / repair-manual / museum-didactic / hospital- discharge still -> classic-archive OR dashboard (pharmacy-insert moved out) - Board section order per chassis: added broadcast-grid row (panes in a flat grid with a ticker above), removed pinboard row - Added note explaining why pinboard was retired — signal to propose a new chassis if zine brands need a non-magazine pattern, not to reintroduce pinboard prototype/reference/design-guide.md: - Chassis list updated alphabetically with broadcast-grid - Prototype section order per chassis: added broadcast-grid (top bar + ticker + hero pane + feed panes + wide event pane + status strip + compact footer), removed pinboard - Cross-chassis rules: CTA-placement guidance for broadcast-grid hero pane; motif placement note includes live-feed panes brand/SKILL.md Phase 3: - Hash mapping reference updated to new alphabetical order brand/reference/brand-profile-schema.md: - chassis value domain updated: 'broadcast-grid' | 'classic-archive' | 'dashboard' | 'magazine'. Note added about v0.3 having pinboard and v0.4 routing zine to magazine Chassis NOT-lists in the three preserved chassis updated: - Replaced "Not a pinboard" line with "Not a broadcast-grid" where useful (classic-archive, dashboard, magazine) After v0.4: 4 chassis, zero anti-toolbox entanglement at the chassis level, coverage of data-register brands that previously had no good structural home. Reference mocks for all four chassis in this repo at tmp/e2e-3/chassis-mocks/{1-classic-archive,2-dashboard,3-magazine, 5-broadcast-grid}.html. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * revert(aem-design/brand): remove chassis system entirely (v0.5) Chassis added in v0.3 and refined in v0.4 (chassis-library shared between brand and prototype, 4 chassis with seed-preference routing) turned out to be a distraction. In the e2e-8/9/10 test runs the chassis did pick deterministically and the boards did render in their chosen chassis's pattern, but the divergence payoff was marginal relative to the surface area of the system — and the cream-ground convergence (the actual tell) happened regardless of which chassis rendered. Removed: - _shared/chassis/broadcast-grid.md · classic-archive.md · dashboard.md · magazine.md (entire chassis library) Reverted to pre-chassis state: - brand/reference/brand-board-template.md — back to a single canonical data contract with section order, design guidelines, and template generation steps. The cream-rebrand ban (vellum, kami, bone, ivory, eggshell) is preserved in the Design guidelines section. - brand/SKILL.md Phase 3 — back to a single render step (read profile → generate HTML → write file → tell designer), no chassis pick, no pick-reason logging. Keeps the explicit "do not default to cream" line that points at toolkit §2.5. - prototype/SKILL.md Phase 2 — removed Step A (chassis pick + provenance block stamping). Phase 2 is now a single render step. - prototype/reference/design-guide.md — removed "Chassis selection" and "Prototype section order per chassis" sections. Replaced the "Cross-chassis rules" block with a "Landing-page section rules" block that holds the same invariants (one primary CTA, briefing drives content, motifs applied within sections, variants share copy + tokens). - brand/reference/brand-profile-schema.md — removed chassis, chassis_pick_reason, and prototype_chassis fields from the _divergence block. Also removed the chassis bullet from the Divergence Tracking notes. What remains from v0.3/v0.4: - _shared/chassis/ directory deleted, but the structural reasoning (data contract vs presentation) has been absorbed into brand-board-template.md as a straightforward single-template spec. - Font decks (toolkit §3) remain — they were always artifact-agnostic and drive real divergence. - Seed system (§2 decade × craft × register) remains. The 'zine register route-to-magazine' routing is moot now that chassis routing is gone. What this does not fix: - Cream ground dominance observed in e2e-8 (justified by seed Embossed leather × Liturgical program), e2e-9 (cream rebranded as "kami" to slip the explicit ban), e2e-10 (cream justified by §2.5 carveout). The toolkit §2.5 Ground-color-by-seed table is the loophole — addressed in the next commit. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/brand): 4th seed dimension + cream rebrand ban (v0.5) Second v0.5 change. Addresses the cream-ground convergence observed across all three e2e-8/9/10 test runs: - e2e-8 (Beumer & Becker, seed Embossed leather × Liturgical program) → cream justified by §2.5 carveout as "liturgical programs print on cream stock" - e2e-9 (Shita-gawa, explicit cream ban in anti-toolbox) → cream rebranded as "kami" (Japanese paper) to slip the ban - e2e-10 (Yadda Dey, seed 1960s × folded-paper × travel brochure, cream on §2.5 "belongs here" row) → cream permitted All three grounds landed at #EDE.../#F0E.../#F5E... — clear cream family. Two changes: 1. Expanded § 1 cream-family entry with rebrand ban and deterministic hex test: - Named rebrands called out: vellum, kami, parchment, washi, bone, oatmeal, eggshell, ivory, biscuit, tan, wheat, fawn, sand, linen, canvas, manila, calfskin, morocco-cream. Renaming cream does not change what it is. - Deterministic HSL-range test: a ground is cream-family when L is 80-97 AND (H is 20°-60° OR S < 15%) AND S < 40%. The LLM can perform this check on its own emitted hex before finalising. - Justification bar narrowed: cream is only acceptable when the brand's primary business is literally printing / publishing / paper-goods / binding / stationery. Adjacent registers (liturgical program, field guide, museum didactic, auction catalogue, travel brochure) are NOT sufficient on their own — they must combine with the brand being in that category. 2. Retired § 2.5 "Ground color by seed" table and replaced it with a new 4th seed dimension: - Ground family · deterministic-random from byte[3] of the seed hash mod 6: cream · stark-white · pale-gray · saturated · dark · monochrome-tint - Stamped in _divergence.seed.ground with seed.picked_by and seed.overrides[] for mixed designer/auto picks - Cream now lands on ~1/6 of no-reference runs by construction, not by template default and not by LLM rationalisation - Designer override available for genuine print/paper brands The retired table was a carveout the LLM used to rationalise cream on most brands. Removing the opportunity for post-hoc rationalisation — by picking the ground BEFORE the LLM reasons about the brand — closes that specific loophole. Schema updates: - brand-profile-schema.md _divergence.seed now includes: - ground field (one of six enum values) - picked_by expanded to include 'mixed' - overrides array for mixed picks - Status banner bumped to v0.5 with a v0.4→v0.5 changelog entry Expected outcome on re-run of e2e-8/9/10 with same brand names/date: - Each brand gets a specific hashed ground family. Cream shows up only when byte[3] mod 6 lands on 0. - For brands that do not hash to cream, cream-family rebrands are actively prevented by the § 1 rebrand list + hex check. - Shita-gawa's "kami" rebrand would now fail both the named-rebrand check AND the deterministic hex test. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(aem-design/brand): cream test uses raw channel warm-bias (v0.5.1) v0.5 rule used 'S < 15' as a proxy for "warm or neutral" in the cream-family hex test. That false-positively flagged pure neutral grays as cream because grays satisfy S = 0. Observed in the Coolors-scraped anchor library: the "Silver Lining" palette (#595959 #7F7F7F #A5A5A5 #CCCCCC #F2F2F2) got classified as cream despite being pure grayscale. v0.5.1 replaces the saturation-as-proxy with raw-channel warm bias: R − B ≥ 5 in the raw 0-255 channels. This still catches every legitimate cream rebrand from the v0.5 list (vellum, kami, oatmeal, bone, eggshell, ivory, parchment, washi, biscuit, tan, wheat, fawn, sand, linen, canvas, manila, calfskin) because those ARE warm at low saturation — their R channel runs higher than their B channel by design. It correctly excludes: - #F5F5F5 (pure gray, R−B=0) - #CCCCCC (pure gray, R−B=0) - #D4E2E8 (cool gray, R<B) - #E8ECEE (cool near-white, R<B) Examples now include R−B deltas for each cream-family example so a reviewer can verify the rule by eye. Status banner bumped v0.5 → v0.5.1 with a short changelog entry. Addresses issue observed when scraping Coolors trending and classifying the 127 extracted palettes — see tmp/e2e-3/tools/ scrape-coolors.py which uses the same rule. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/brand): palette picker + library integration (v0.6) Add an offline palette-selection phase to the brand skill. Replaces the weakest link in the pipeline — LLM-invented color palettes — with a pick-from-library flow whose colors come from real Coolors trending palettes. Every hex carries a source URL back to coolors.co. New files (127 palette JSONs + 4 structural files): _shared/palettes/ library.json consolidated index of all 127 entries cream/ (5) warm near-white grounds stark-white/ (6) true-white, near-achromatic pale-gray/ (5) cool near-neutral saturated/ (78) bold, colored-light or colored-dominant dark/ (20) ink or deep-jewel grounds monochrome-tint/ (13) tinted near-neutral with loud accent Each palette file: source URL, scraped_at, name, 5 hexes, classified anchor candidate, classification (ground_family + hue_bias + saturation_level + energy + cream-flag list). Scraped from coolors.co/palettes/trending on 2026-04-24 via tools/scrape-coolors.py in the session repo. _shared/palette-picker.md classifier vocabulary + filter scoring - §1 keyword dictionaries for 5 descriptor dimensions (energy, contrast, saturation_level, hue_bias, ground_family) - §2 filter scoring rules (exact/loose hue match, ±diff saturation + energy, ground-family 100-point match) - §3 HSL + raw-channel warm-bias helpers (for library regeneration) - §4 UI contract pointing at the template - §5 designer corrections section brand/reference/pick-ui-template.md - full HTML template with {{placeholders}} - dark-background layout, numbered cards 1-5, gold border on recommended pick, cream-flag indicator per swatch, Coolors source links, designer instruction footer Modified: brand/SKILL.md - NEW Phase 2 Palette Selection (119 lines) Step A derive description from designer brief OR synthesize from voice/pillars/seed Step B classify via palette-picker.md §1 keyword vocabulary Step C filter and score palettes/ per §2 rules Step D render pick UI at aem-design/_palette-pick.html Step E wait for designer pick (integer 1-5, palette name, or refine) Step F record in in-memory profile + _divergence.palette_source - Renumbered downstream phases: Design Personality 2→3, Render Brand Board 3→4, Approval Gate 4→5 - Updated "No .impeccable.md" cross-reference to point at Phase 3 brand/reference/brand-profile-schema.md - Added _divergence.palette_source object: method enum, library_version, description_used, classification block, candidates_shown array, recommended_index, picked_index, picked_palette_name, picked_palette_source - New "palette_source sub-block" doc section under Divergence Tracking - Backwards-compatible: older profiles without palette_source still read valid Zero external dependencies. Entire flow runs offline from bundled data. Expected outcome on a no-reference run: - brand skill reaches Phase 2 - if designer provides a palette brief, runs classifier; else synthesizes one from voice + seed - writes aem-design/_palette-pick.html with 5 candidate cards - designer picks a number - brand profile's colors.primary[] gets the Coolors hexes with brand-native role names from the designer (or derived if absent) - _divergence.palette_source records the full chain for audit Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/prototype): Phase 0b imagery mode · generated or placeholder Extend Phase 0 (previously "Variant Count") into "Setup" with a second question: whether the prototypes should use branded placeholder rectangles (default) or real generated images. When generated, the skill collects the model name and credential source. Changes: prototype/SKILL.md Phase 0 renamed "Variant Count" → "Setup": - Step 0a (variant count) — existing content preserved - Step 0b (imagery mode) — NEW Step 0b asks the designer: "For images in the prototype, do you want: (a) branded placeholder rectangles — fast, offline, free, or (b) real generated images?" If (b), two follow-ups: 1. Which model? Options presented with provider + env-var name: - gemini → Gemini 3 Pro Image Preview · GOOGLE_API_KEY - flux → fal.ai FLUX Schnell · FAL_KEY - imagen → Vertex Imagen 3 · GOOGLE_APPLICATION_CREDENTIALS - dalle → OpenAI DALL-E 3 · OPENAI_API_KEY 2. Where's the credential? - "env" → read from the named env var directly - A .env file path → read the var from that file (key stays in memory) - A raw token → used only for the session, not written to disk Imagery mode recorded in each prototype's provenance block: <!-- aem-design:provenance imagery_mode: placeholder | generated imagery_provider: gemini | flux | imagen | dalle | null imagery_credential_source: "env" | "<file-path>" | "<session-only>" imagery_generated_at: <ISO timestamp> (only when mode=generated) images: [ { section, path, prompt_hash } ] --> Never logs the actual key value anywhere. Re-ask-once logic: on subsequent invocations in the same project, the skill reads the previous prototype's provenance block and re-uses the recorded mode; only re-asks if missing. Phase 2 consequences: - mode=placeholder → existing branded-rectangle path per design-guide.md - mode=generated → invoke ai-image-generator skill (from eds-site-builder, sumi, or testing plugin — whichever is registered) with the briefing's # Imagery direction + brand photography rules + target aspect ratio. Images at aem-design/prototypes/images/{page}-{section}.png - Cache by prompt_hash — iteration doesn't re-generate unchanged prompts - Fallback when generation fails: retry once simplified, then fall back to placeholder for that slot only, stamp imagery_fallback: true, never abort the whole page over one failed image design-guide.md Imagery section rewritten with three named paths: (a) briefing source hint — highest priority (b) branded placeholder (default) (c) generated image via ai-image-generator soft-dep Alt-text requirement preserved across all three paths; "image" or "photo" as the whole alt is explicitly disallowed. Zero new dependencies. The generation path uses ai-image-generator which is already available in the user's environment via three installed plugins. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(aem-design/prototype): default variant count 1 → 2, require confirm Phase 0a used to default to 1 variant when the designer skipped. v0.7 changes the default to 2 (a pair to compare) and asks for explicit confirmation: "I'll produce 2 variants of each prototype so you can compare — confirm, or give a different number (1-4)?" Valid responses: - "ok" / "confirm" / "yes" / Enter / "2" → 2 variants (default) - 1 → single variant - 3 or 4 → that many variants - other → re-ask once Rationale: the multi-variant comparison workflow (Phase 3 table, Phase 4 iteration on the chosen variant) is the more-useful default for design decisions. A single variant skips the compare step; a pair captures the most value for least effort. Pipeline automation also defaults to 2 now — no explicit variant count in the prompt means the skill renders 2 and presents them. No downstream changes: Phase 2 rendering logic already handled both single-variant and multi-variant modes. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * refactor(stardust): rename aem-design plugin to stardust Move from plugins/aem/design/ to plugins/stardust/ (top-level, sibling of plugins/aem/, plugins/app-builder/, plugins/creative-cloud/). Rename plugin identifier to "stardust". Rename content-folder convention aem-design/ to stardust/ inside user projects. Broaden scope language from AEM-EDS-specific to platform-agnostic static-HTML design. No backward compatibility — clean break. Filesystem moves (git mv, rename history preserved): - plugins/aem/design/ → plugins/stardust/ - plugins/stardust/skills/aem-design/ → plugins/stardust/skills/stardust/ - plugins/stardust/evals/aem-design-navigator/ → plugins/stardust/evals/stardust-navigator/ Manifest updates: - .claude-plugin/marketplace.json — entry name aem-design → stardust, source path updated, description broadened to "static-HTML design projects" from "AEM Edge Delivery Services sites" - plugins/stardust/.claude-plugin/plugin.json — name + description + keywords (removed aem, eds; added palette, wireframes) - plugins/stardust/tile.json — adobe/aem-design → adobe/stardust, summary broadened, skill map key aem-design → stardust - plugins/stardust/README.md — full rewrite with broadened scope, added "What makes stardust different" section covering the v0.2-v0.7 palette library + divergence toolkit + seed system Navigator skill (skills/stardust/SKILL.md): - Frontmatter name aem-design → stardust - Full prose rewrite via word-boundary sed Source updates (single pass of \baem-design\b → stardust across all .md and .json files in the plugin): - ~60 invocation-path refs (/aem-design:* → /stardust:*) - ~100 content-folder refs (aem-design/ → stardust/) - ~20 prose mentions - 1 eval-folder name ref Scope-broadening edits (beyond the sed): - prototype/SKILL.md description: "no EDS or dev-server dependency" → "no build system or dev-server dependency" - prototype/SKILL.md: "EDS-free" → "platform-free", "EDS-independent" → "platform-independent" - brand/SKILL.md icons/ note: reframed as "platform-agnostic" with EDS as one example - wireframes-guide.md: "not an EDS block type" → "not a downstream block/component type"; "concrete EDS site" → "concrete site" AEM/EDS still mentioned in scope-correct contexts as ONE EXAMPLE of a downstream target (alongside "another SSG", "a different CMS", "a framework component library") — those are kept. Eval criterion IDs like "no_eds_references" kept stable so eval history still runs. BREAKING: existing user projects with aem-design/ artifact folders are not found by /stardust. Users must `mv aem-design stardust` in each project. Installed plugin cache at ~/.claude/plugins/cache/adobe-skills/aem-design/0.1.0/ becomes stale; re-install from the marketplace as stardust to replace. Validation: all 5 skills valid under new paths. Zero remaining aem-design references. Commit log preserves rename history via git-mv + git similarity-detection. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(stardust): open HTML artifacts in browser after every write Every HTML artifact the stardust skills produce now opens in the designer's default browser immediately after it's written. No more "tell the designer to open the file" wording — the skill invokes the platform-appropriate open command via Bash. Rule added to _shared/skill-contract.md: ## Opening HTML artifacts - macOS: open "<absolute-path>" - Linux: xdg-open "<absolute-path>" & - Windows: start "" "<absolute-path>" - Skill detects platform from environment, picks matching command. - Unknown platform → try open, fall back to xdg-open. - Open after EVERY write (including iteration re-writes). - Multi-variant mode: open all variants simultaneously for tab- compare. - Pipeline-automation mode (no designer present): skip the open. - Bash-unavailable environments: fall back to telling the designer the path. Skill updates: - brand/SKILL.md: - Phase 2 Step D (palette pick UI): `open stardust/_palette-pick.html` immediately after write. Designer message updated from "open the file" to "the UI is already open." - Phase 4 (brand board): `open stardust/brand-board.html` immediately after write. SLICC browser-panel path still works. - wireframes/SKILL.md Phase 3 (Serve): - Open each emitted wireframe in the browser immediately. - Multi-page run: open each for tab-compare. - Old code-block with `open` + `xdg-open` side-by-side replaced with clear "skill does the open, not the designer" rule. - prototype/SKILL.md Phase 3 (Serve): - Open each emitted prototype in the browser immediately after Phase 2 render. - Multi-variant mode: open all variants simultaneously so the designer can flip between browser tabs for the comparison-table evaluation. - `python3 -m http.server` path preserved for cases that need a real HTTP origin (fetch/service-workers), but file:// via `open` is the primary path (covers ~95% of prototype needs). - prototype/SKILL.md Phase 4 (Iterate): - Every iteration re-opens the file so the designer sees the change immediately. Browsers with live-reload pick up the change on re-open. Rationale: the designer-opens-manually pattern added friction at every gate. Phase 2 palette pick, Phase 4 brand board, wireframe review, and prototype variant comparison all required the same mental gear-shift of copy-path-then-open-browser. The open is one Bash call; putting it in the skill removes a persistent small annoyance. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |