사용 방법
- 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
- 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
- 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/draftbit/mobile-taste-skill --skill mobile-redesign-app시작 전 확인
- 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
- 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.
mobile-redesign-app: The Whole App, In Phases
Method files, read in full first: ../_shared/references/redesign-protocol.md, ../_shared/references/review-rubric.md, ../_shared/references/tells.md, and ../_shared/references/navigation.md (for the nav audit).
Flow
1. Full audit
Stack sniff, then the complete review-rubric pass (static + live when reachable) across every screen in the route tree, PLUS:
- Nav audit: the current route tree diffed against the navigation.md mechanical
checks (tab count/logic, container choices, back behavior, deep links). Emit the current tree as a Nav Read so problems are visible in the same format the fix will use.
- Token audit: where the design system lives, or the finding that it doesn't.
- Dial reading of the app as-built; platform posture consistency.
- Write
design-baseline.json- the redesign will be measured against it.
2. Mode + scope agreement
Preserve vs overhaul (ask once). Then present the plan as PHASES, each one lever from redesign-protocol.md applied app-wide:
Phase 1 Tokens + type/spacing (touches theme files + every screen mechanically)
Phase 2 Color onto semantic tokens; dark mode
Phase 3 States (skeleton/empty/error/offline/refresh) per data screen
Phase 4 Press feedback + motion layer
Phase 5 Screen recomposition (worst screens first, from the findings)
Phase 6 Navigation restructure (ONLY if the nav audit demands it - see below)Per phase: findings closed, files touched, what will not change. Get approval for the plan once, then checkpoint (short summary + before/afters) at each phase boundary rather than re-asking.
3. Navigation restructure (the gated phase)
This skill may PROPOSE nav changes - as a current-vs-proposed Nav Read diff with every never-change-silently item flagged (route names, deep links, tab order, learned gestures) and a migration note (redirects for renamed deep links, state migration). It ships ONLY after explicit approval of that diff. If navigation is fine, say so and skip the phase; most redesigns should.
4. Execution
The redesign-protocol fix loop, per phase: atomic commits, before/after pairs, findings closed in the baseline, risk governor at app scale (stop > 20%, hard cap 40 fixes per run - a large app takes multiple runs, and that is the design, not a limitation). Phase order is not negotiable mid-run; a tempting Phase-5 recomposition discovered during Phase 2 becomes a finding, not a detour.
5. Close out
Regression run of the rubric → delta table against the starting baseline (this is the proof the redesign worked), remaining findings, refreshed MOBILE-DESIGN.md (reads, dials, token locations, Nav Read) so future sessions inherit the new system.
Hard boundaries
- Overhaul mode changes visual language, never content or IA without the gated diff.
- Accessibility wins from the audit are regression-tested, not assumed.
- Every phase's screens must pass the mobile-taste Section 14 pre-flight before the
phase checkpoint counts as done.
