yazzang-homelab/replica-suite

Makes a clone legally and visibly its own and ships it - name candidates with trademark/domain/store checks, new palette and logo brief, voice, a sweep that finds leftovers of the original, landing page, pricing, store listing lint, preflight checklist, production setup, DNS, monitoring and mobile release. Use for "브랜딩", "이름 짓기", "리브랜딩", "배포", "출시", "스토어 등록", "원본 흔적 검사", "rebrand the clone", "name my app", "make it mine", "deploy it", "ship it", "launch checklist", "app store listing", after app-parity-test says shippable.

Implements a clean-room clone from the recon map and app-spec - shell and foundation first, a vertical slice of the core flow, then screen-by-screen build from written specs (optionally parallel builders), with on-device features working for real and backend parts mocked or wired. Use for "클론 구현", "앱 만들어줘", "화면별 구현", "재구현", "build the clone", "implement screen S07", "rebuild this app in Expo/Next", "start building from the spec", after app-spec.

Evidence-driven high-fidelity rebuild of a web application or creative/animated site with quantitative gates - scope fingerprint, immutable mirror with sha256 ledger, offline run, line-traceable reverse notes, milestone port with deviation register, pixel/numeric parity gates, cold-head audit, source-readable output. Use for "1:1 복제", "사이트 복원", "정밀 재현", "미러링 후 재구현", "인터랙션/애니메이션 정확히 재현", "내 사이트 마이그레이션", "죽은 사이트 복구", "1:1 rebuild", "pixel-accurate rebuild", "forensic rebuild", "archive and restore my site".

Router for whole-application analysis and clean-room re-implementation skills (recon, spec, build, parity/QA, brand+deploy, binary/APK/Electron internals, forensic high-fidelity rebuild). Use when the user says "앱 클론", "앱 복제", "앱 재구현", "앱 분석 어디서부터", "clone an app", "rebuild X", "reverse engineer an app", "which app-* skill", or when unsure which app- skill applies.

Evidence-based inspection of shipped application artifacts - native executables, .NET/JVM, Electron/ASAR, APK/IPA, Python-packaged apps, firmware, installers, web bundles, HAR captures - to explain features, compare versions, recover architecture and feed a clean-room spec. Uses the REA MCP/CLI when available, otherwise standard tools. Use for "바이너리 분석", "APK 분석", "일렉트론 앱 분석", "디컴파일", "리버스 엔지니어링", "버전 비교", "내부 구조 파악", "reverse engineer this app", "what does this binary do", "analyze this APK/asar/exe", "compare two builds", "REA".

Verifies a clone against its spec and the original - test plan from flows, Playwright/e2e and console-clean route sweep, bug log by severity, weighted feature-parity score, colour-agnostic screenshot layout diff, iterative visual-diff loop and design review. Use for "클론 검증", "패리티 점수", "원본과 비교", "QA", "버그 찾기", "화면 비교", "test my clone", "how close is it to the original", "what's missing", "visual diff", "parity check", "is it shippable", after app-build.

First step of whole-application re-implementation. Maps an app (web, iOS, Android, desktop) into screens, flows, components, inferred data model and user-pain evidence from public sources, the user's own account, or a folder of screenshots. Use for "앱 클론", "앱 복제", "앱 분석", "이 앱 어떻게 동작해", "화면 목록 정리", "clone this app", "replicate X", "reverse engineer the UX of", "map the screens of", "what do users hate about X", "should I build an alternative to X".

Turns an app-recon map into a buildable specification - stack choice, database schema, API/route list, design tokens (measured system, new identity), component specs and a thin-slice build order. Use for "아키텍처 설계", "스택 선정", "스키마 설계", "디자인 토큰", "기능 매트릭스 확정", "plan the clone", "design the database", "write the schema", "what stack for my version of X", "design system from screenshots", after app-recon.

Static analysis of Android APK/XAPK/AAB/JAR/AAR - fingerprint the framework, decompile with jadx/apktool/Fernflower, read manifest and BuildConfig, recover obfuscated Kotlin class names, trace UI-to-network call flows and extract HTTP endpoints (Retrofit, OkHttp, Ktor, Apollo, Volley) into a catalog. Use for "APK 분석", "안드로이드 앱 디컴파일", "jadx", "apktool", "안드로이드 리버스", "코틀린 난독화 복원", "앱 API 엔드포인트 추출", "decompile this APK", "reverse engineer Android app", "find API endpoints in app", "Kotlin name recovery", "split APK / XAPK". Feeds mobile-api-extraction and app-spec.

Turn mobile-app evidence (decompiled Android/iOS output, JS bundles, HAR/proxy captures) into a deduplicated endpoint catalog, then an OpenAPI 3 / API spec for a clean-room reimplementation, bridging to app-spec. Use for "앱 API 정리", "엔드포인트 카탈로그", "OpenAPI 스펙 작성", "모바일 앱 백엔드 재구현", "API 명세 역추출", "extract app endpoints", "build API spec from APK/IPA", "endpoint inventory to openapi", "document the app's backend", "HAR to spec".

Flutter/Dart AOT 앱 분석, libapp.so 및 libflutter.so 조사, 플랫폼 채널과 상태 흐름 복원. Analyze Flutter binaries and produce evidence-backed specifications for migration or reimplementation.

Frida 런타임 관측, Android Java·native 또는 iOS Objective-C 호출 흐름 확인과 재현 가능한 동적 분석. Trace app behavior with version-matched Frida instrumentation and evidence-ledger handoff.

모바일 앱 이식·재제작을 위한 Android APK, iOS IPA, Flutter, Unity, Frida 및 API 분석 분류 진입점. Route mobile reverse engineering and reconstruction requests by runtime and evidence.

Static analysis of iOS apps - unpack IPA/.app/Mach-O/.dylib/.framework, inspect Info.plist, entitlements, ATS and URL schemes, dump Objective-C/Swift classes (ipsw class-dump, otool, rizin, Ghidra headless), check FairPlay encryption, trace call flows and extract HTTP endpoints (URLSession, Alamofire, AFNetworking, GraphQL, WebSocket). Use for "IPA 분석", "iOS 앱 리버스", "아이폰 앱 디컴파일", "Mach-O 분석", "class-dump", "ipsw", "otool", "Info.plist entitlements 확인", "iOS 앱 API 엔드포인트 추출", "reverse engineer iOS app", "analyze this IPA", "decrypt check cryptid". Feeds mobile-api-extraction and app-spec.

Unity IL2CPP 앱 분석, global-metadata.dat와 네이티브 바이너리 연결, 타입·메서드 복원 및 Ghidra/IDA 검증. Recover evidence-backed Unity behavior specifications for game ports.

Analyze an existing (retro) game binary or ROM, extract its assets and rules, and drive a documented, verified port or reimplementation to the web (HTML canvas). Use for: game reverse engineering, ROM/EXE disassembly with labels, sprite/tile/palette extraction, data table decoding, scriptable emulator for validation, web port of a DOS/NES/Game Boy/Amiga game, REVERSE.md project loop. Korean triggers: 게임 리버스, 롬 분석, 에셋 추출, 스프라이트 타일 팔레트 추출, 웹 포팅, 게임 재구현, 레트로 게임 복원, 에뮬레이터 대조 검증.

Drive Ghidra for native reverse engineering headlessly (analyzeHeadless, PyGhidra, a ghidra-rpc style daemon, or Ghidra MCP servers) with an iterative checkpoint-apply-cascade-verify loop and the SourceType trust model so the agent never corroborates its own guesses; includes pointers for IDAPython/IDALib and radare2/rabin2 when Ghidra is not the tool. Use for "Ghidra", "analyzeHeadless", "PyGhidra", "rename functions/retype structs in Ghidra", "decompile and annotate", "Ghidra MCP", "IDAPython", "idalib", "radare2 r2 script", "patch diff two binaries", "고드라/기드라 분석", "헤드리스 디컴파일", "함수 이름 적용", "구조체 복구", "IDA 스크립트", "radare2 사용".

Router for the native- category: reverse engineering compiled binaries, Ghidra/IDA/radare2 workflows, runtime tracing, C/C++ reconstruction, retro game porting, and front-end JS reverse. Use first for: 바이너리 리버싱, 네이티브 분석, 디컴파일, 고스트라 자동화, 함수 복원, 게임 롬 분석, JS 서명 분석, reverse engineer this binary, which reverse-engineering skill should I use, ELF/PE/Mach-O triage, Ghidra MCP, trace runtime, rebuild C from binary.

Front-end JavaScript reverse engineering with browser-debugger tooling: locate how a web app builds request signatures/encrypted parameters, trace XHR/fetch/WebSocket initiators, sample runtime values at minimal intrusion, rebuild the algorithm in Node with evidence-driven environment patching, then optionally deobfuscate. Use for: signature chain, sign/token/encrypted param analysis, js-reverse MCP, breakpoint on XHR, env patching (补环境), first divergence. Korean triggers: 프론트엔드 JS 리버싱, 서명 파라미터 분석, 암호화 파라미터 추적, XHR 브레이크포인트, 런타임 샘플링, Node 환경 복원, 난독화 해제.

Rebuild readable, compilable C/C++ from a compiled binary (ELF/PE/Mach-O/firmware) function by function, then verify the rebuild against the original's behaviour with independent checking, structural diffing, build/test gates and a parity grade. Use for: decompile to source, recreate function in C, matching decompilation, reconstruct class from Ghidra output, reimplement binary module, reverser/checker loop, parity check. Korean triggers: 바이너리 C/C++ 재구성, 디컴파일 결과를 소스로 복원, 함수 재작성, 원본 동작과 동일한지 검증, 클래스 복원, 패리티 검증.

Trace values, callers, initialization, dispatch and runtime boundaries in native software with a resumable, evidence-ledger investigation - static provenance first, sandboxed runtime observation only for specific unresolved questions, static/runtime reconciliation, build-to-build comparison, and checkpoints that survive context compaction or agent switches. Use for "where does this value come from", "who writes/calls this", "trace this function at runtime", "indirect call target", "compare two builds / patch diff", "resume the investigation", "frida/gdb/strace trace", "값 추적", "호출자 추적", "런타임 트레이싱", "빌드 비교", "조사 이어서 하기", "분석 노트 체크포인트".

Evidence-first first-contact triage of a native binary (ELF/PE/Mach-O/firmware/shared library) before any deep reading - identify format/arch, harvest free names (symbols, RTTI, debug info, build paths), classify strings with their referrers, read the import table as an attack-surface/capability map, fold away known library code, recover symbols for stripped functions, and decide decompiler vs disassembly per question. Use for "analyze this binary", "what does this exe/dll/so do", "stripped binary", "recover function names", "decompiler output looks wrong", "바이너리 분석", "리버싱 시작", "스트립된 바이너리 심볼 복구", "문자열/임포트 분석", "디컴파일 vs 디스어셈블리".

프로그램·사이트 이식과 재제작 통합 진입점. Route app analysis, website cloning, screenshot-to-code, mobile package analysis and native reconstruction through evidence, specification, implementation and verification gates.

Analyze a UI screenshot, mockup, Figma export or phone screenshot BEFORE writing code - fix canvas/DPR, extract design tokens (color, type, spacing, radius, shadow), layout regions, component inventory, exact text, asset roles, and write a build spec. Use for "스크린샷 분석", "디자인 토큰 추출", "레이아웃 분석", "이미지에서 색상/폰트/간격 뽑아줘", "analyze this screenshot", "extract design tokens from image", "screenshot to spec", or as step 1 of any shot-to-* skill.

Render a built UI, screenshot it at the target viewport/DPR, compare it to the source image region by region, and iterate fixes until differences are cosmetic - with scripts for screenshots, state capture, interaction audit, diff heat maps and multi-viewport QA. Use for "스크린샷 비교", "픽셀 비교 반복 수정", "시안과 구현 차이 확인", "렌더링 비교 루프", "visual diff loop", "compare my render to the target image", "fix visual differences", "parity check".

Design-first pipeline from a brief or reference image to a built site - write a locked blueprint, generate a UI mock-up with an image tool, get confirmation, export separate clean assets, then implement and QA against the approved mock-up (optional deploy). Use for "디자인 시안 만들고 구현", "레퍼런스 이미지로 웹사이트 제작", "목업 생성 후 코드화", "랜딩 페이지 시안부터", "design mockup to website", "reference image to site", "generate UI design then build it", "blueprint then build".

Router for the screenshot/image-to-UI-code skill family (shot-*) - pick the right skill for turning a screenshot, mockup, phone app capture or design brief into HTML, React/Vue/Svelte, UnoCSS pages, React Native, Qt QML, or a generated design pipeline. Use for "스크린샷 코드 변환 어디서 시작", "이미지로 UI 만들기", "시안 구현 스킬 추천", "which skill for screenshot to code", "image to UI index", "shot skills overview".

Rebuild a UI screenshot, mockup or HTML/CSS/Tailwind page as a native Qt Quick (QML) project for PySide6 and/or C++/CMake - measured tokens, a Theme singleton, QtQuick.Templates controls, headless render-and-diff verification, kiosk/embedded profiles. Use for "스크린샷을 QML로", "Qt Quick UI 구현", "PySide6 화면 복원", "임베디드 HMI 시안 구현", "screenshot to QML", "image to Qt Quick", "convert HTML to QML", "Raspberry Pi kiosk UI from mockup".

Rebuild a mobile app UI from a folder of phone screenshots (screens and states) as a runnable Expo / React Native project (Expo Router + NativeWind) - cluster screens and states, infer navigation, write per-screen specs, build, run the render-diff loop on Expo web, and make on-device interactions real. Use for "앱 스크린샷으로 React Native 앱 만들기", "모바일 화면 캡처를 Expo로 복원", "앱 UI 클론", "screenshots to React Native", "clone this mobile app UI", "rebuild app from screenshots", "Expo from mockups".

High-fidelity, section-by-section conversion of segmented design images (page slices, landing/marketing/dashboard mockups) into UnoCSS-styled Vue or Astro page code inside an existing project, with a mandatory pre-implementation brief, asset fulfilment plan and per-section screenshot diff. Use for "UnoCSS로 시안 구현", "디자인 이미지를 Vue/Astro 페이지로", "섹션별 고충실도 복원", "랜딩 페이지 시안 코드화", "design image to UnoCSS page", "section-driven high fidelity", "pixel-faithful Vue page from mockups".

Turn a UI screenshot, mockup or Figma export into working web code - single-file HTML/CSS (plain or Tailwind), React (Vite/Next.js), Vue 3 or Svelte - matching the RENDERED result through a render-compare-refine loop. Use for "스크린샷을 HTML로", "이미지 디자인을 웹페이지로 복제", "시안 코드로 만들어줘", "React/Vue/Svelte로 변환", "screenshot to HTML", "image to React", "clone this design", "recreate this page from the image".

Rebuild a captured website as clean React/Next/Vite code by building the foundation first, then dispatching one builder agent per section/sub-component in parallel worktrees from spec files, merging, assembling, and driving a gated repair loop to a ready-for-review handoff. Use for "웹사이트 클론 빌드", "섹션별 병렬 구현", "병렬 에이전트로 사이트 복제", "build the clone", "rebuild this page section by section", "parallel builders".

Capture trustworthy evidence from a live website before cloning it - raw vs rendered HTML, full-page and per-section screenshots, fonts, assets, JS chunks, fingerprints, multi-page crawl. Use for "웹사이트 캡처", "사이트 분석 시작", "레퍼런스 수집", "capture this page", "recon a website", "grab assets/fonts/screenshots before rebuilding", or as step 1 of any website clone.

Turn a live website into a reusable DESIGN.md covering palette, typography, components, layout, motion, responsive rules and agent prompts, plus an HTML preview using DOM/computed-style evidence. Use for "DESIGN.md 만들어줘", "사이트 디자인 시스템 추출", "스타일 가이드 문서화", "look and feel 분석", "extract design tokens from this site", "write a style guide from URL" — documentation, not rebuilding the page.

Turn a captured website into exact design facts - computed-style dumps per section, design tokens, fonts, SVG/icons, hover/scroll/click state diffs, animation and interaction-model analysis, responsive breakpoints, and one component spec file per section. Use for "CSS 추출", "디자인 토큰 뽑기", "애니메이션 분석", "스크롤 인터랙션 분석", "extract computed styles", "reverse-engineer animations", "write component specs".

Analyze a live website for cloning using Firecrawl (sitemap map + rendered scrape) plus Playwright (computed styles, animations, assets, screenshots) and produce a research pack with page brief, tokens, motion and asset manifests for builders. Use for "사이트 구조 분석", "파이어크롤로 사이트 맵 뜨고 분석", "클론 전에 리서치", "Firecrawl Playwright analysis", "analyze this site before cloning", "reverse-engineer page structure and animations".

Router for the web- category — analyzing, capturing, rebuilding, mirroring, verifying and documenting existing websites. Use first for "웹사이트 클론", "사이트 복제", "웹 분석", "디자인 시스템 추출", "픽셀 비교", "WebGL 사이트", "clone this website", "replicate a web page", "which web skill should I use".

Mirror a public page into a locally runnable static replica (HTML, same-origin assets, nested CSS urls, fonts, media), inventory every replaceable resource, then brand-swap it for the user's own product (copy, logo, colors, fonts, media, meta, links) and prepare it for safe publication. Use for "웹사이트 미러링", "로컬 복제 실행", "브랜드 교체", "리브랜딩", "mirror this page locally", "swap the brand/logo/copy", "static replica", "리소스 목록".

Independent, read-only QA review of a finished website clone by a separate reviewer agent, ending in a strict PASS/BLOCK verdict with P0/P1/P2 findings. Use for "독립 QA", "클론 검수", "리뷰어 에이전트로 검증", "최종 승인 전 체크", "independent reviewer", "release gate for the clone", "don't trust the builder".

Prove a website clone matches its reference with a multi-breakpoint screenshot matrix, pixel diff, runtime health checks and an iterative auto-fix loop. Use for "픽셀 퍼펙트 검증", "스크린샷 비교", "클론 차이 찾아서 고쳐줘", "pixel diff", "visual regression", "does my clone match", "fix the differences until it matches".

Clone or test a WebGL / Three.js / Rive / canvas-heavy / scroll-scrubbed website by detecting the runtime, mirroring its resources locally (models, textures, wasm, workers, shaders), serving it, and running multi-viewport runtime QA. Use for "WebGL 사이트 복제", "Three.js 3D 사이트 클론", "캔버스 애니메이션 그대로", "Rive 애니메이션", "runtime mirror", "my screenshot clone lost the 3D scene".