skills/design-taste-web/SKILL.md
Generation-time design taste for web UI work. Anti-cliche bans, layout and motion hard rules, and client-intent dials. Advisory only - shapes drafts; declared measured contracts remain the sole gate authority.
npx skillsauth add astro44/Autonom8-Agents design-taste-webInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Loaded for: dev_design, dev_implement on UI/web platforms.
Purpose: raise first-draft visual quality so fewer closeout repairs are
needed. Read the brief, set the dials, then execute within the hard rules.
design.taste_dials:
design_variance: 5 # 1 = strict/symmetric/centered … 10 = experimental/asymmetric
motion_intensity: 4 # 1 = static, hover-only … 10 = cinematic, scroll-driven
visual_density: 5 # 1 = gallery-sparse … 10 = dashboard-dense
State a one-line design direction before writing any code (audience + tone + one distinctive move). Every layout/motion choice should be explainable from the brief and the dials.
design_variance > 4. Centered is a
choice for manifesto-style briefs, not a reflex.visual_density: sparse briefs get generous whitespace and
large type; dense briefs get tight, aligned, information-rich modules —
but empty regions are never "design", they are absence.motion_intensity; at ≤ 3 use only micro-feedback (hover,
focus, pressed).prefers-reduced-motion whenever intensity > 3.Before returning, verify: no banned cliché present; hero fits viewport with ≤ 2 headline lines; no repeated layout family; grids fully populated; type scale consistent; contrast margins comfortable; every animation justified; copy free of em-dashes, emoji, and placeholder text.
This checklist is self-discipline, not proof. The worker's rendered-surface contract will measure the page; design so the measurements are comfortable, not barely passing.
development
Scores proposal complexity against codebase surface. Uses proposal text analysis and readiness stats to determine decomposition tier and agent count.
testing
Fast filesystem readiness scan — counts docs, source files, manifests, platform signals. Produces initial ReadinessReport for agent spawning decisions.
testing
Merges bookend agent reports into revised readiness, complexity, and decomposition plan. Produces the final evidence-backed assessment consumed by sprint-architect-agent.
development
Rigorously reasons about definitions, proofs, and computations in algebra, analysis, discrete math, probability, linear algebra, and applied math. Verifies derivations, spots invalid steps, and states assumptions clearly. Use when solving or proving math problems, reviewing mathematical arguments, modeling with equations, interpreting statistics, or when the user mentions proofs, lemmas, theorems, integrals, series, matrices, optimization, or numerical methods.