Design, review, diagnose, or implement user-driven interface behavior—taps, presses, drags, swipes, scrolling, paging, transitions, gesture conflicts, momentum, springs, interruption, cancellation, boundaries, feedback, and discoverability—especially for Apple platforms. Use when an interaction feels laggy, stiff, abrupt, disconnected, hard to cancel, or hard to discover. Never invent numeric motion values. Do not use for static visual styling or decorative animation without an interaction problem.
Review a prepared, normalized snapshot report of the current Apple guidance used by Cupertino Taste. Use only when explicitly invoked by the repository maintenance routine. Never fetch sources, browse, edit files, or change dates; return the required JSON decision for deterministic validation and application.
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Cupertino Taste provides Claude Code design skills distilled from selected WWDC design sessions.
Distilled by WeZZard in China, Assembled from WWDC.
design-fluid-interface designs, reviews, diagnoses, and implements user-driven interaction behavior. It is based on “Designing Fluid Interfaces,” WWDC 2018 session 803, with historical examples qualified by current Apple Human Interface Guidelines.
Use it for taps, presses, drags, swipes, scrolling, transitions, gesture conflicts, momentum, springs, cancellation, boundaries, feedback, and discoverability. A static screenshot receives an affordance review; claims about fluidity require a recording, runnable prototype, or real-device use.
maintain-current-apple-guidance is an explicit contributor maintenance skill. It reviews a prepared report of normalized Apple guidance and returns a schema-defined decision. It never activates implicitly, fetches sources, edits files, or chooses a date. The maintenance scripts invoke it through Codex headless mode only when normalized source content changes.
Raw transcripts, videos, slides, screenshots, and downloaded Apple material are not included. The private research corpus lives outside this repository on DAS.
The maintenance routine makes acquisition, source validation, normalization, hashing, comparison, receipt validation, and date updates deterministic. Codex supplies the semantic judgment only for changed normalized content. Its result must follow a JSON Schema and match the exact report and source hashes, but model inference is not bit-for-bit deterministic.
Run one of these four commands from the repository root:
Fetch, normalize, compare, and write a private report:
./skills/maintain-current-apple-guidance/scripts/check-current-apple-guidance
Review a changed report through the explicit skill:
./skills/maintain-current-apple-guidance/scripts/review-current-apple-guidance --report /absolute/path/to/report.json
Accept a validated report and advance the baseline and Last checked date:
./skills/maintain-current-apple-guidance/scripts/accept-current-apple-guidance --report /absolute/path/to/report.json
Run the composed check, conditional review, and acceptance routine:
./skills/maintain-current-apple-guidance/scripts/run-current-apple-guidance-routine
The scripts archive responses, normalized snapshots, reports, Codex events, and receipts outside Git. When DAS is mounted, the default private archive is /Volumes/DAS/3.Resources/Documentations/WWDC/cupertino-taste/current-apple-guidance. Set CUPERTINO_TASTE_GUIDANCE_ARCHIVE to override it. Without DAS, the scripts use the platform cache directory.
The composed command skips Codex when normalized content is unchanged. Byte-level representation changes that normalize to the accepted content also skip Codex. An intentional source-manifest change produces a changed evidence set and requires review.
Expected routine outcomes use stable exit codes:
| Code | Meaning |
|---|---|
0 | Sources are current, or review, acceptance, or the composed routine completed successfully. |
10 | The check found changed normalized content or a changed source manifest. Review is required. |
20 | A source fetch failed. |
21 | A fetched source failed its identity or format contract. |
22 | An input, manifest, baseline, option, or routine state is invalid. |
23 | A report, snapshot, structured review, receipt, or target state failed validation. |
24 | Codex is unavailable, failed, timed out, or returned a nonzero status. |
The repository does not schedule this routine through CI or launchd. Contributors run it explicitly when they want to verify current guidance.
Load the plugin from the repository root:
claude --plugin-dir /Users/wezzard/Artifacts/Repositories/com.github/WeZZard/cupertino-taste
Validate its plugin structure:
claude plugin validate --strict /Users/wezzard/Artifacts/Repositories/com.github/WeZZard/cupertino-taste
The plugin can also be installed from the WeZZard marketplace after its tagged release is registered there.
The skill publishes original synthesis and links to the canonical WWDC 2018 session and current Apple design guidance. It does not distribute Apple transcripts or media.
Cupertino Taste is an independent project. It has not been authorized, sponsored, or otherwise approved by Apple Inc. Apple and WWDC are trademarks of Apple Inc.
The MIT license applies only to original material in this repository. It does not grant rights to Apple content or third-party source material.
npx claudepluginhub wezzard/skills --plugin cupertino-tasteWeZZard's Claude Code skills collection
Tools for auditing and improving Claude Code skill definitions
Capture and apply human-ruled knowledge the model cannot internalize — user preferences and output-style rulings, settled by oracle: research the world, ask the human, experiment on the unknown.
Delegate tasks to external agent CLIs through one router and one brief contract — probed facts before use, write isolation through worktrees, a multi-model audit panel, and concurrent image generation with failure memory.
Memory compression system for Claude Code - persist context across sessions
Frontend design taste skills including brutalist, minimalist, soft, redesign, stitch, and more
Unified capability management center for Skills, Agents, and Commands.
45% cost reduction measured. Cache expiry prevention, SubTask auto-delegation, zero-cost context restoration, real-time cost dashboard. The only Claude Code plugin built from CC source analysis.
Ultra-compressed communication mode. Cuts 65% of output tokens (measured) while keeping full technical accuracy by speaking like a caveman.
Frontend design skill for UI/UX implementation