The comp rules composition; the world rules material. Adds the authority
ranking to new-work.md, rewrite and token discipline to the craft floor,
token and motion evidence checks to the finish reviewer, derived text
inks to the questionnaire seed, and stale-seed disclosure at attended
stops. From the ekihana run's confirmed drift.
Co-authored-by: Cursor <cursoragent@cursor.com>
Name Cursor browser_navigate as the in-IDE viewport, then other in-app tools, then the system opener. Same ladder in serve-question stdout so agents stop using macOS open while the IDE browser works.
AI-assisted.
Co-authored-by: Cursor <cursoragent@cursor.com>
On the questionnaire path, chat now asks only assets, named references, and one anti-reference; color, type, and motion stay in the picker. Interview-only still asks all five. visual-cues.md drops the retired Q-keys and re-anchors font pairs from surfaces, references, and assets.
AI-assisted.
Co-authored-by: Cursor <cursoragent@cursor.com>
CJK and other non-Latin families render the picker's previews and scale sheets wrong until language-aware pairs exist; the fallback deck comment matches the new visual-cues Step 6 rule.
AI-assisted.
Co-authored-by: Cursor <cursoragent@cursor.com>
Direction rolls on projects without DESIGN.md now refuse until the user accepts document --seed or re-runs with --seed-declined carrying their actual skip answer; IMPECCABLE_SEED_DECLINED=1 remains the unattended harness escape. Hide the design context shell when [hidden] so it does not reserve viewport space. Extend concept-seed tests for the seed gate.
AI-assisted.
Co-authored-by: Cursor <cursoragent@cursor.com>
When new-work reaches world creation and DESIGN.md is missing, pause and
offer /impeccable document --seed (skippable) so the visual world can come
from the user's own choices instead of an assigned roll.
AI-assisted change (Cursor agent), directed by the maintainer.
Co-authored-by: Cursor <cursoragent@cursor.com>
The picker staleness check added on this branch had replaced main's
dry-run branch, so `--dry-run` executed the real build. Keep the picker
hash check but only on a real run, matching main's dry-run contract.
AI-assisted (agent-implemented, maintainer-directed).
Co-authored-by: Cursor <cursoragent@cursor.com>
Add gallery avatars, tighten the experience index layout, and sync the improved previews gallery replica.
Co-authored-by: Cursor <cursoragent@cursor.com>
Tighten Landing, App UI, and Docs type-board copy, color, spacing, and gallery experiments so the improved preview gallery matches the live picker.
AI-assisted commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
Board-stage read artboards get thicker bar weight in picker.css; the gallery replica adds matching experiments plus desktop caption circles for Portfolio restrained/committed/drenched and drenched color roles on cell 21.
AI-assisted commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
Boost docs strategy-stage line and dot weight to match App UI legibility, keep phone artboard lines proportional via pv-text scaling, and extend the gallery replica experiments for cells 14–18.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add sequential labels to figcaptions and iframe titles in the gallery builder, then regenerate baseline and improved HTML.
Co-authored-by: Cursor <cursoragent@cursor.com>
Un-ignore tmp/questionaire for the gallery toolchain, baseline HTML, and the improved replica with drenched, committed, and full-palette color fixes.
Co-authored-by: Cursor <cursoragent@cursor.com>
Mirror the screen 02 neutral contrast badge on the strategy strip, share alert
painting between both strips, and render richer PRODUCT.md context in the
design document with matching visual-cues schema guidance.
AI-assisted commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
Replace the neutral palette hint warning with an on-swatch badge, simplify the finish loader screen, and keep typography/material vignettes ink-only until hover.
Co-authored-by: Cursor <cursoragent@cursor.com>
Product, Brand, Typography, and Material vignettes now use anim-join so their gold accent reads at 25% opacity before hover and brightens on hover, matching the Color tile.
Co-authored-by: Cursor <cursoragent@cursor.com>
Type scale and layout return to the linear path before configure; the hub shrinks to five cards with updated navigation. Also fixes palette tooltip clipping, drops the design-context masthead and icon hub notes, and locks the mosaic shell to the viewport without overflow.
Co-authored-by: Cursor <cursoragent@cursor.com>
Pills shrink to content width and use slightly more padding so values read as compact chips rather than full-width rows.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add the intro lede, per-row patina answer marks with DOM-sourced notes, mode-tile hover, and tighter card typography including inline pills and title arrows.
Co-authored-by: Cursor <cursoragent@cursor.com>
Let questionnaire runs finish early or jump into the seven remaining questions from a summary card grid, with hub navigation and per-answer summaries.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add neutralContrastIssue() with a 7:1 ink floor and a 3:1 primary
separation floor, re-run it from renderPreview(), and paint the palette
hint in vermilion while the working neutral is bad. The warning clears
as soon as the neutral is fixed.
Prepared with AI assistance (Cursor).
Co-authored-by: Cursor <cursoragent@cursor.com>
Each option now draws its own page rather than a shared field with a
different column count. Simple grid holds strict paired rows at equal
image measures, balanced alternates wide images against captions, and
freeform crops a lead, adds an upper portrait, and spills the card row
into a set of prints with their own captions and marks. All three carry
handset sequences at the reference measures.
Behind them: one shared alignment-grid component mounted on every body
that asks for it, code-native nav primitives placed at the reference
measures rather than left to their grids, a project image-tone contract
the boards key their fills on, and footer link groups.
Prepared with AI assistance (Cursor).
Co-authored-by: Cursor <cursoragent@cursor.com>
Abdul Wahab
