BRAND IDENTITY
MAINBuilt between shifts. — 교대 사이에 지었다. 인생이 보이는 한 줄.
SUBYou had the tools. I had a phone. — 대비로 긴장 만드는 서브 카피.
CTA28 systems. Zero permission. — 팩트 + 도발. 증거 섹션/버튼 문구.
POSITIONING개인이 운영하는 R&D 출판 엔진 — 내부 정체성 선언. README, CLAUDE.md, Manual용.
This is not a brand built to sell. It's a brand built by surviving.
팔려고 만든 브랜드가 아니다. 살아남아서 생긴 브랜드다.
5 LOCKED CATEGORIES
HARD RULE
Every task must belong to one of these 5. If it doesn't fit → do not build.
"I am not a developer who builds features. I am a person who repeatedly produces systems."
PHLPhilosophy / Language
Thought → executable structure. Constitution layer. Other categories must not violate this.
papyrus · CLAUDE.md (31) · 00_TRUTH (33)
whitepaper / spec / prompt protocol
AUDIOAudio Engine
Not music apps. Audio production pipeline — conducting, rendering, batch processing.
parksy-audio · apk-lab/audio-tools · local DAW
wav/mp3 + render log
VISUALVisual Engine
Not design. Assembly, arrangement, automation — visual production line.
parksy-image · tango-magenta
assets + scene sheet
ACTIONAction Engine
Not building UI. Bypassing OS to automate actions — overlays, captures, device setup.
apk-lab (8 APK) · cloud-appstore (8 PWA) · phoneparis/baptism
apk/pwa + automation log
INFRAInfrastructure
Not a code repo. A company OS — Git-based, serverless, replicable.
28 repos · 72 Actions · 3 broadcast stations · GitHub Pages
dashboard + repo map + action runs
JUDGMENT FLOW
New task comes in →
Ask: "Is this PHL / Audio / Visual / Action / Infra?"
Fits one → PROCEED
Unclear → DEFER
Fits none → DO NOT BUILD
EVIDENCE RULE
EVERY TASK COMPLETION REQUIRES
Link 1 — page / release / video URL (publicly reachable)
Log 1 — short .md with date, category, what was done, what was produced
No link + no log = task never happened. This is the audit trail.
SCOPE BOUNDARY (RED LINE)
DO NOT CROSS
SaaS로 바로 팔지 않는다 — No direct SaaS sales
패키징/배포는 외주/파트너로 넘긴다 — Packaging & distribution = partners
나는 엔진/판정/증빙이 역할이다 — My role = engine + judgment + evidence
This line keeps Parksy at architect, not developer.
EVALUATION FRAME — 이 시스템을 평가하는 올바른 잣대
WHAT THIS IS
1인 R&D 스튜디오 + 실행형 출판 시스템
Output = 서비스가 아니라 판정 가능한 기록물 (whitepaper, spec, dashboard, evidence)
Revenue source = 제품 판매가 아니라 판정 능력 + 구조 설계력
Role = 파는 사람이 아니라 의사결정에 개입하는 아키텍트
WRONG FRAME — 이 잣대로 평가하면 오진한다
Stripe/결제 있냐? → Scope 밖. 결손 아님.
GA/SEO/CRM 있냐? → Scope 밖. 결손 아님.
대중 트래픽 스케일 되냐? → 목표 자체가 아님.
SaaS IR 피칭 가능하냐? → 게임 자체가 다름.
Scope Boundary로 선언한 항목을 결핍으로 채점하는 것은
평가자의 기준 통제 실패지, 대상의 부족이 아니다.
CORRECT FRAME — R&D + 출판 스튜디오 기준
연구 방법론 명시 — Category Lock + Evidence Rule + Judgment Flow 9/10
재현 가능성 — Fork Guide로 28 repo 복원 가능 9/10
문서화 완결성 — CLAUDE.md 31개 + whitepaper + manual 9/10
증거 추적 — 4,731 커밋 + Evidence Whitepaper 감사 완료 8.5/10
스코프 명확성 — 5 category + scope boundary + red line 9.5/10
독창적 기여 — 핸드폰 1대 + Termux + AI → 28 repo OS. 선례 없음 9/10
INTENTIONAL EXCLUSIONS — 없는 게 아니라 안 한 것
EXCLUDED 결제 시스템 (Stripe, 멤버십) → SaaS 안 팔기로 선언됨
EXCLUDED SEO / GA / CRM → 대중 트래픽이 목표가 아님
EXCLUDED 소셜 자동화 / 뉴스레터 → 마케팅은 스코프 밖
EXCLUDED 비디오 편집 도구 → 제로 편집이 설계 원칙
이 항목들이 "빠져있다"고 지적하는 건
채식주의자한테 "고기를 안 먹는다"고 감점하는 것과 같다.
GENUINE NEXT STEPS — 진짜 다음 단계
NEXT 브랜치 5명 실사용 검증 — 구조는 있다. 사람이 아직 안 움직였다. 설계 완료 → 실험 미수행. 이건 결함이 아니라 다음 단계다.
NEXT 외부 발표 / 피어리뷰 — R&D라도 발표 전이면 미검증. 서랍에 있으면 원고지 출판이 아니다. 내보내야 출판이다.
NEXT 에셋 파이프라인 실물 투입 — 파이프라인은 있는데 물건이 아직 비어있다. 첫 번째 실물 에셋이 통과해야 검증.
이 3개는 Scope 밖이 아니다. Scope 안인데 아직 안 한 것이다.
"의도적 배제"와 "아직 안 한 것"을 구분하는 게 정직한 평가다.
OPERATION SCENARIOS
Tap a scenario to expand steps.
INFRA Start a new Claude Code session
1Open Termux on Galaxy
2Run claude — CLAUDE.md auto-loads with Category Lock
3State your task. AI will auto-judge which category it belongs to.
4If it fits 5 categories → proceed. If not → AI tells you to stop.
cd ~ && claude
PHL Write a new operational document
1Navigate to papyrus repo
2Tell Claude what to document (whitepaper, policy, spec)
3Claude writes to docs/ AND hq/ (both copies)
4Review → commit → push → live on GitHub Pages
cd ~/dtslib-papyrus && claude
ACTION Publish a new article on eae.kr
1Open Claude in eae.kr repo — enters Publisher Mode automatically
2Say: category + text + URL (optional YouTube link)
3Claude creates MDX file in src/content/{category}/
4Commit → push → GitHub Actions → live on eae.kr
cd ~/eae.kr && claude
RULE
eae-blueprint category requires YouTube link. No standalone publish.
ACTION Build/update an APK
1Navigate to apk-lab repo
2Tell Claude which app to modify (8 apps in apps/)
3Claude edits Dart code → commit → push
4GitHub Actions builds APK → available on store page
cd ~/dtslib-apk-lab && claude
ACTION Deploy/update a PWA
1Navigate to cloud-appstore repo
2Edit the target PWA in its directory (8 apps at root level)
3Commit → push → Vercel auto-deploys
cd ~/dtslib-cloud-appstore && claude
AUDIO Run audio processing pipeline
1parksy-audio/local-agent/ has 6 API endpoints
2On GPU PC: cd local-agent && python server.py
3Available: audio_cut, extract_midi, clean_midi, merge_midi, render, gradient_mp4
4Or run Flutter app (parksy-audio-tools) on phone for capture
cd ~/parksy-audio && claude
VISUAL Run image automation scripts
1parksy-image/scripts/ has 11 Python files
2On GPU PC: resize, CAD gen (DXF/SCAD/SVG), batch_upload, assemble
3Output goes to assets/ → manifest.json auto-updates via Action
cd ~/parksy-image && claude
INFRA Sync all repos at once
1Use the CLI tools in ~/bin/
2dts status — check all repos
3dts pull — pull all
4dts push — push all
dts status
계열사
INFRA Onboard a new branch member
1Clone dtslib-branch template structure
2Create new repo: {name} with CLAUDE.md + 00_TRUTH + core/
3Add to dtslib-branch registry
4Existing branches: koosy, gohsy, artrew, papafly, namone
INFRA Set up a new device (baptism)
1Clone phoneparis repo to new device
2Run tools/baptism/install.sh
3Select profile and scenario
4Termux + Claude Code + git = ready
cd ~/phoneparis/tools/baptism && bash install.sh
ESSENTIAL COMMANDS
Session Start
cd ~ && claude
Opens Claude Code. ~/CLAUDE.md auto-loads with Category Lock.
Repo-Specific Session
cd ~/dtslib-papyrus && claude
cd ~/eae.kr && claude
cd ~/dtslib-apk-lab && claude
cd ~/parksy-audio && claude
cd ~/parksy-image && claude
Each repo has its own CLAUDE.md with repo-specific instructions.
Fleet Management
dts status
dts pull
dts push
dts sync
계열사
Fleet-wide git operations across all repos.
Evidence Audit
for d in ~/*/.git; do r=$(dirname $d); echo "$(basename $r): $(cd $r && git rev-list --count HEAD)"; done
Count commits per repo. Used for Evidence Whitepaper verification.
Line Count
find ~/parksy-image -name "*.py" | xargs wc -l
find ~/dtslib-apk-lab/apps -name "*.dart" | xargs wc -l
Measure code volume for evidence tracking.
FORK & OPERATE GUIDE
If you fork the entire dtslib1979 GitHub account, here's how to get operational.
Step 1: Device Setup
Required: Android phone (Galaxy S20+)
Install: Termux → pkg install git nodejs python
Install: Claude Code CLI (Max plan required)
pkg install git nodejs-lts python
Step 2: Clone All Repos
Clone the 28 repos to home directory. Each contains CLAUDE.md (AI SOP) and 00_TRUTH (identity).
gh repo list dtslib1979 --limit 50 --json name -q '.[].name' | xargs -I{} gh repo clone dtslib1979/{}
Step 3: Verify Structure
Check that CLAUDE.md and 00_TRUTH exist across repos.
find ~/ -maxdepth 2 -name "CLAUDE.md" | wc -l
find ~/ -maxdepth 2 -name "00_TRUTH" -type d | wc -l
Step 4: First Session
Run claude from home directory.
~/CLAUDE.md loads → Category Lock active → AI knows the system.
Say your first task. AI auto-judges category.
cd ~ && claude
Step 5: Understand the 5 Categories
PHL — Design rules, whitepapers, operational docs
Audio — Audio pipeline, MIDI, rendering
Visual — Image automation, CAD, webtoon
Action — APK/PWA apps, device automation
Infra — Git OS, Actions, deployment
REMEMBER
This system is not about building features.
It's about repeatedly producing systems.
Every task → Category judgment → Execute or defer.
No exceptions.