Score UI design quality and auto-fix until it passes gate

Scores a UI file's design quality 0-100 against StyleSeed's design language, with a per-category breakdown, the worst offenders cited by line number, and a prioritized fix list ordered by score gain - a quantified counterpart to /ss-review's prose audit: /ss-review says what's wrong, /ss-score says how good it is overall and what to fix first.

Scoring runs across seven weighted categories summing to 100: Color discipline (18, reading DESIGN-LANGUAGE Section 1/18/72 and VISUAL-CRAFT C4), Hierarchy & typography (18, Sections 2/3/4/16 plus the Font Size table and VISUAL-CRAFT C2), Layout & rhythm (14, Sections 13/14/15/61 and VISUAL-CRAFT C1), Cards & elevation (12, Sections 7/8/12/1 and VISUAL-CRAFT C3), States & a11y (18, Sections 11/70/71/72 and VISUAL-CRAFT C3), Motion & interaction (8, Sections 24/59 and engine/motion), and Coherence (12, VISUAL-CRAFT C0's "one choice per axis" rule). Each category starts at full marks and loses points for specific, cited violations - e.g. color discipline deducts for pure black text/backgrounds, more than one decorative accent hue, emoji used as UI icons, or status conveyed by color alone; states & a11y deducts for a missing empty/loading/error state, contrast below 4.5:1, touch targets under 44px, or no visible focus ring; coherence deducts for mixed radius personalities, competing accent hues, or mixed icon styles - flagged as the category that most predicts a UI "looks AI-generated," since it weights system-wide consistency over per-component prettiness. Every category is clamped at 0 before summing.

### Design Score: 70 / 100   (src/app/Dashboard.tsx)

████████████████░░░░░░  C-

Color discipline      13/18   ▓▓▓░  #000 headings (l.12,40); orange+blue+green accents (l.28-34)
Hierarchy & typography 15/18  ▓▓▓▓  number/unit 1:1 on hero (l.18)
Layout & rhythm        11/14  ▓▓▓░  two identical KPI rows (l.22-31)
Cards & elevation       8/12  ▓▓░░  1px borders doing separation (l.22)
States & a11y          11/18  ▓▓░░  no empty/loading state; focus ring missing (l.55)
Motion & interaction    6/8   ▓▓▓░  default fade, not a named seed
Coherence               6/12  ▓▓░░  sharp cards (l.22) + pill buttons (l.48); 3 accent hues (§C0)

### Fix first (highest score gain)
1. Add empty + loading states to the orders list       → +7 states (§71)
2. Unify radius (pick soft 8-12px) + collapse to one accent → +9 coherence+color (§C0, §2)
3. Drop the 1px borders, use tone + ≤8% shadow         → +4 cards  (§7)

Re-score after: ~92 / 100.

Scores map to letter bands (90+ A, 80-89 B, 70-79 C, 60-69 D, below 60 F). For a directory, it prints one score line per file, then the full breakdown for the lowest-scoring file.

In Gate mode, /ss-score runs as a loop rather than a one-off: score the just-generated UI, and if below 80, apply the "fix first" list (via /ss-review) and re-score, repeating up to about three times or until reaching 80. The 80 bar is a floor, not a ceiling - stop once it's met rather than chasing a perfect 100, since the goal is keeping an obviously-incoherent first draft (rainbow status colors, emoji icons, multiple accents, missing states) from ever reaching the user, not polishing an already-clean UI further.

Use it to get a quantified design-quality number with a prioritized fix list, or as an automated quality gate before showing generated UI to a user. For a quick pass/fail check before committing, use /ss-lint instead; for a full prose audit with fixes, use /ss-review; it does not apply fixes itself in plain scoring mode, and scoring is meaningless for non-UI files (logic, config).

Input is a UI file or directory of files to score, read directly from source (never guessed). Output is a total score with letter grade, a per-category breakdown citing specific line numbers, and a fix list ordered by expected score gain.

Reads violation rules from StyleSeed's own DESIGN-LANGUAGE.md and VISUAL-CRAFT sections and engine/motion, and hands off fixes to /ss-review (prose audit) and implicitly /ss-motion when running in Gate mode's fix-and-re-score loop.

Teams that want UI design quality tracked like test coverage - a single comparable number per file, cited evidence for every deduction, and an automatable quality gate that blocks obviously incoherent first-draft UI from reaching users.