Product Design
SkillMediaSet the visual goal for a major UI or flow change, then implement and verify the rendered result. Use for new pages, navigation, or redesigns. Do not use for tiny copy, spacing, or color fixes.
Then ask your AI: use the Product Design skill
Details
Instructions available. Your AI can read the instructions. Execution depends on the setup they require.
Account requirements not reviewed. Check the skill instructions before use; Ahel provides instructions and does not run this skill.
No other account needed.
Add it as a step in a pipeline you build on Ahel, or connect Ahel to Claude, ChatGPT, Cursor, Claude Code or Codex and ask your AI to use it.
What this skill tells your AI
The instructions your AI receives, as published by milksu-official/milksu in skills/product-design/SKILL.md and read by Ahel’s review.
Turn an interface request into one clear visual target, a working implementation, and rendered evidence.
Establish the product context
- Read repository instructions and inspect the current route, neighboring screens, shared components, tokens, and existing user flow.
- Capture the current rendered screen and every supplied reference before proposing a direction. If a required reference cannot be opened, stop before inventing a replacement.
- State the user's primary task, the information that must remain visible, and the one primary action. Remove blocks that do not help that task.
- Reuse the product's design language and real assets. Do not introduce a second component system, decorative dashboard cards, fake data, placeholder capability, handcrafted icons, or a new visual style without an explicit request.
Select the visual target
For a new page, major redesign, or changed user journey, create a concrete visual mock before editing product code. Show hierarchy, layout, controls, expanded states, and realistic content. Ask for confirmation when the visual direction materially changes the product.
For a small visual defect that preserves the existing design, skip the mock and use the current rendered screen as the target.
When implementing from a screenshot, match the source rather than improving it from memory. Record only necessary differences caused by the host product, accessibility, or viewport behavior.
Implement the confirmed flow
- Inventory existing behaviors before replacing markup.
- Compose shared components and tokens. Keep the information hierarchy shallow and the primary action obvious without explanatory paragraphs.
- Make navigation, menus, forms, filters, toggles, loading, empty, disabled, and error states required by the main journey work with realistic data.
- Preserve keyboard access, visible focus, labels, readable density, and supported narrow layouts.
- Do not deploy or publish unless the user explicitly requests it.
Verify the rendered result
Use the repository's canonical tests and a real preview or packaged App. For web surfaces, also follow frontend-visual-qa when available.
Compare the target and implementation at the same viewport and state. Check spacing, clipping, typography, borders, responsive behavior, interaction states, Console errors, and failed requests. A build or source review alone is not visual proof.
Report the confirmed target, implemented journey, tested states, rendered evidence, intentional differences, and any remaining uncertainty.
Signals
- GitHub stars
- 25
- Forks
- 6
- Last commit
- Oct 2026
Advanced
- Item type
- skill
- Key
product-design-milksu-official- Source
- github.com/milksu-official/milksu
github.com/milksu-official/milksu