Design rigorfor AI coding

Research web design principles and real examples to connect a brand-fit direction with responsive implementation and rendered-screen verification.

npx designpaca
See the deliverables Explore example sites
The designpaca process, assembled as a seven-stage design compiler Seven aligned plates turn a brief into references, direction, tokens, implementation, preflight checks, and a verified result.
Brief · References · Direction · Tokens · Build · Preflight · Proof

Decisions carry through to code and checks

Principles selected from research are recorded, translated into design tokens and implementation rules, then verified through repeatable checks.

Explore example sites

Three designpaca output specimens

Direction

  1. Structure
  2. Tone
  3. Detail

Evidence converges into one visual position.

References are separated by role before their decisions are combined.

Tokens & type

Surface
--surface
Ink
--ink
Signal
--accent

Type, color roles, and spacing become reusable constraints.

Verification contract

  1. Line count
  2. Content inset
  3. Contrast
  4. Keyboard path

The same rules are measured at narrow, medium, and wide widths before completion.

Connect judgment to implementation in seven stages

Each stage records the decisions the next one needs, keeping the build anchored to an intentional direction.

  1. 0

    Brief gate

    Set the goal, audience, required features, language, and constraints before design work begins.

  2. 1

    Reference research

    Study real examples that solved a similar problem, then separate useful principles from habits to avoid.

  3. 2

    Direction

    Choose the visual direction, information hierarchy, and accepted risk to establish one coherent design language.

  4. 3

    Design tokens and budget

    Define color, type, spacing, motion, and performance limits as concrete values.

  5. 4

    Implementation

    Turn the chosen principles into responsive layouts, materials, interactions, and accessible states.

  6. 5

    Preflight audit

    Check viewport behavior, contrast, keyboard access, metadata, tokens, and performance with repeatable tests.

  7. 6

    Decision record

    Document accepted and rejected choices so the same quality standard carries into the next build.

Verify the result and record the evidence

Run responsive behavior, accessibility, token hygiene, and metadata through the same repeatable checks.

7 Pipeline stages
Connect research, contextual judgment, implementation, verification, and decision records from stage 0 through stage 6.
Evidence · context Criteria kept distinct
Apply universal usability criteria separately from style decisions that fit the project.
Rendered screens Two kinds of review
Automate functional checks, then separately review the composition and hierarchy users actually see.
View the audit checks
  1. Make sure core content stays intact without clipping or overlap at 320px.
  2. Check that heading, body, and button spacing follows the token system.
  3. Measure contrast for text and interface controls.
  4. Complete navigation and primary actions with a keyboard.
  5. Keep content and primary actions available with reduced motion and at 200% zoom.
  6. Review real before-and-after screens for composition, hierarchy, and responsive behavior.
  7. Connect sources, metadata, language, alt text, and ARIA labels to the rendered page, then check tokens, SEO, and unnecessary code.

Connect it to the AI coding tool you already use

The interactive installer detects your workspace and places the skill files where supported tools expect them.

npx designpaca
  • Codex, Claude Code, Cursor, and other AI coding tools
  • User-wide or project-scoped installation
  • Updates, install diagnostics, and manifest-based removal