Design product interface
Guides a research-to-release workflow for interactive product interfaces across web applications, macOS, Windows and Linux desktop, iOS and Android, and cross-platform apps. Use when creating or substantially redesigning an application, dashboard, admin or operational tool, settings surface, or multi-step product workflow; use design-and-build-website instead for conversion-oriented marketing websites.From its SKILL.md
npx -y skills add K95M65/AI_ONBOARD --skill design-product-interfaceAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 28 days oldThe repository was created 28 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
8.6 KB, ~1.6k tokens by cl100k_base, as published. Nobody here has run it
Design a product interface
Take an interactive product from explicit user work to a platform-appropriate, verified release. Treat the target platform as part of the product definition, not as a skin applied after designing a generic interface.
Establish the operating context
- Read repository instructions and inspect the product, current workflows, data model, UI architecture, design system, platform APIs, tests, analytics, support material, and release path.
- Confirm that the task is an interactive product. For a marketing website whose main job is persuasion
and conversion, use
design-and-build-website. - Identify every supported platform, form factor, input method, window model, offline requirement, accessibility target, and distribution channel.
- Decide whether the work is:
- native-first: optimize for one operating system and its conventions;
- shared product, adapted surfaces: preserve concepts and workflows while intentionally varying navigation, commands, layout, and controls by platform;
- cross-platform shell: share implementation without pretending the platforms are identical.
- Ask only for missing facts that would materially change the product model. State reversible assumptions when autonomous progress is authorized.
Use conduct-user-research when important user, environment, workflow, or usability claims need evidence.
Use shape-product-opportunity when the desired outcome, opportunity, assumptions, or experiment boundary
is still unclear.
Load only the relevant platform guidance
- For browser-delivered products, read references/platform-web-app.md.
- For macOS products, read references/platform-macos.md.
- For iOS or Android products, read references/platform-mobile.md.
- For Windows, Linux, Electron, Tauri, Qt, or other desktop products, read references/platform-desktop.md.
- For a multi-platform product, read every applicable platform reference. Define one shared product model, then record deliberate platform deltas. Do not settle for a lowest-common-denominator interface.
Treat these references as durable heuristics. When exact platform behavior, APIs, store requirements, or accessibility rules matter, check the current official platform guidance.
Run the workflow
1. Define the product role and strategy
Read references/strategy-and-workflows.md. Produce a compact brief covering users, jobs, environment, task frequency, stakes, domain language, objects, permissions, success measures, platforms, constraints, and non-goals. Use assets/product-interface-brief-template.md when a durable artifact will help.
Do not design screens until the primary users, work, and consequences of error are explicit.
For a first-run, adoption, or stalled-user problem, read references/activation-and-onboarding.md. For a product whose core behavior uses generative or agentic AI, read references/ai-powered-experiences.md.
2. Model workflows, structure, context, and density
Read references/layout-context-and-density.md. Define the conceptual model, information architecture, task flows, navigation, layout paradigm, command model, content hierarchy, scope cues, and density. Start with the hardest representative workflow, not a decorative home screen.
For a table paired with a chart, map, diagram, timeline, canvas, or another representation of the same data,
use coordinate-data-views.
3. Specify components, content, and states
Read references/components-and-states.md. Define component contracts, domain-language labels, complete interaction states, validation, recovery, permissions, loading, empty content, long content, and destructive-action behavior. Use realistic data shaped by the real schema.
Use design-product-content for a substantial terminology, onboarding, error, help, or localization
problem. Use evolve-design-system when the task changes shared tokens, component contracts, governance,
versioning, or migration rather than merely consuming the existing system.
4. Establish the visual system and build vertical slices
Derive tokens and visual language from the product role, platform conventions, brand, content density, and use environment. Preserve sound existing patterns. Use native controls and system behaviors when they improve familiarity, accessibility, or integration; customize deliberately when product identity or task performance justifies it.
Use define-brand-foundation when brand strategy or voice is missing. For native products, read
references/native-visual-and-motion.md before establishing visual
direction.
Implement in this order:
- Data and state model for the chosen workflow.
- Application shell, navigation, commands, focus, and platform integration.
- Reusable primitives and one complete representative vertical slice.
- Remaining workflows and secondary surfaces.
- Loading, empty, error, partial, success, offline, permission, and recovery states.
Render or run each meaningful slice before multiplying the pattern.
5. Test and refine on the target platforms
Read references/validation-and-release.md. Run repository checks and exercise the actual application with representative data, input methods, window or viewport sizes, system settings, permissions, and failure conditions. Validate task completion and recovery, not merely visual similarity.
Use audit-accessibility for a dedicated cross-platform audit and remediation plan. Use
measure-product-experiments when success requires an event contract, baseline, experiment, or post-release
learning loop.
6. Review, export, and release
Use an independent design-review agent after UI changes and a verifier agent for final acceptance when those roles are available. Add security review for authentication, authorization, sensitive data, payments, uploads, external input, privileged operations, or cross-account administration. Use an accessibility-review agent when accessibility behavior is material, after remediation, or before a high-impact release.
Follow the repository's established packaging, signing, notarization, store, deployment, or distribution path. Obtain authorization before a live release unless the user explicitly requested it. Smoke-test the released artifact or environment and report the version, target, evidence, limitations, and rollback path.
Compose with specialists
- Let native UI, frontend implementation, component scaffolding, data visualization, browser automation, security, and deployment skills own their bounded mechanics.
- Use
develop-apple-platform-appfor Swift, SwiftUI, AppKit, Apple persistence, testing, profiling, packaging, signing, or distribution work. - Use platform SDK guidance that matches the repository rather than adding a framework merely because it is familiar.
- Preserve accepted artifacts and decisions across phases. Reopen them only when implementation or testing provides new evidence.
Apply quality rules
- Optimize for the user's work, not the number of visible features.
- Use the user's domain vocabulary and make action scope unmistakable.
- Match information density to expertise, frequency, screen size, and consequence of error.
- Make keyboard, pointer, touch, assistive technology, and automation behavior appropriate to the target.
- Keep repeated components aligned through stable slots and shared tokens.
- Expose commands through the platform's expected discovery paths.
- Design recovery alongside success.
- Never claim completion from a successful compile or attractive static screen alone.
Finish with evidence
Report the product model, platform decisions, workflows and components delivered, checks and task scenarios run, accessibility and independent-review status, release state, and remaining product or operational dependencies.
What ships with it: 13 files
31.6 KB alongside SKILL.md
agents/
- openai.yaml286 B
assets/
references/
- activation-and-onboarding.md2.3 KB
- ai-powered-experiences.md2.3 KB
- components-and-states.md4.4 KB
- layout-context-and-density.md4.2 KB
- native-visual-and-motion.md2.1 KB
- platform-desktop.md2.1 KB
- platform-macos.md3.1 KB
- platform-mobile.md2.0 KB
- platform-web-app.md1.9 KB
- strategy-and-workflows.md3.1 KB
- validation-and-release.md2.7 KB