Materializr
오픈소스 파라메트릭 3D CAD로 제약 스케치·솔리드 모델·나사·SVG·텍스트 조각·STL/STEP 내보내기를 지원합니다
아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
프로젝트 폴더에서 /nomorevibe verify를 실행하면 소유권을 확인하고 소개를 직접 관리할 수 있습니다.
1분 만에 확인하기A design method for AI coding tools: routes, skills and agent profiles for Claude Code, Codex and OpenCode.
저장소 README에서 가져옴
README 전체를 불러오는 중입니다.
README 전문 보기 ↗JADO (Just Another Design Orchestrator) is a design method for AI coding tools. It routes a design request to one of eight routes and gives the agent the procedures, profiles and review steps for that workflow. Claude Code, Codex and OpenCode each get native agent profiles and skills.
This repository contains the method sources, a command-line installer and the playbook site that documents every workflow with a Mermaid diagram.
Clone the repository and install its dependencies:
git clone https://github.com/mattiamichini-design/jado.git
cd jado
npm ci
Then install JADO for your user:
npm run jado -- install --global
The command asks which tools to set up, prints the plan and waits for confirmation. To choose the tools up front, pass --agents:
npm run jado -- install --global --agents claude-code,codex,opencode
Keep the checkout after installing. Updates and refreshes run from it.
Start a new session in your AI tool afterwards. Tools load profiles and skills when a session starts, so an open session keeps its old instructions.
Ask for design work as usual: a screen, a flow, a review, alternatives to compare. The instruction block that JADO adds tells the agent to read ~/.jado/instructions.md and pick a route. You can also name the entry point, jado-run, or ask ask-jado which skill or workflow fits.
Each run has three settings: stage (explore, refine, ship), involvement (autopilot, checkpoints, pair) and budget (lean, normal, deep). Set project defaults with a line in the project instructions, for example JADO defaults: stage=refine involvement=checkpoints budget=normal, or change one during a run with jado: autopilot.
To print a route's plan without running it:
npm run jado:plan -- --workflow R05 --stage explore --budget lean
jado/workflows.json defines ten workflows: the router, eight routes and the setup workflow.
| ID | Route |
|---|---|
00-router | Entry and route selection |
R01 | Patch |
R03 | Decide with the user |
R04 | Build |
R05 | Explore alternatives |
R07 | Staged delivery |
R08 | Council |
R09 | VariantHub on explicit request |
visual-research | Patterns and style research |
setup-project | Install or link JADO |
The playbook pages under /docs/jado describe each one. jado/README.md collects all the diagrams in one file.
| Path | Content |
|---|---|
~/.jado/ | The method: graph, policies, runtime packets, catalog, JADO skills, upstream skills, instructions.md, migration.md and the install.json manifest |
~/.claude/agents/jado-agent-*.md | Claude Code profiles |
~/.claude/skills/ | Links to JADO and upstream skills |
~/.claude/CLAUDE.md | A managed instruction block |
~/.codex/agents/jado-agent-*.toml | Codex profiles |
~/.codex/config.toml | A managed block that registers the Codex profiles |
~/.codex/AGENTS.md | A managed instruction block, written to AGENTS.override.md instead when that file exists |
~/.config/opencode/agents/jado-agent-*.md | OpenCode profiles |
~/.config/opencode/AGENTS.md | A managed instruction block |
~/.agents/skills/ | Links to JADO and upstream skills, read by Codex and OpenCode |
Each row applies only to a tool you selected. The managed block sits between # BEGIN JADO — managed design entry and # END JADO — managed design entry, and the rest of each file stays as it was. A skill with the same name as a JADO skill is replaced; other skills in those folders are left alone.
The installer also removes the skills it manages from ~/.agents/.skill-lock.json, so the skills CLI does not update them a second time.
The paths follow CLAUDE_CONFIG_DIR, CODEX_HOME, XDG_CONFIG_HOME and XDG_STATE_HOME when those point inside your home directory. To try the installer without touching your home directory, add --user-root DIR or --dry-run.
The profiles set a model for each tool, such as claude-opus-5-5 for Claude Code and gpt-6.1-sol for Codex. They are defaults: edit the installed profile if your account uses other models. OpenCode profiles use the session model.
npm run jado -- uninstall --global
Uninstall reads ~/.jado/install.json, prints its plan and asks for confirmation. It disables a scheduled refresh, deletes the files JADO wrote that are unchanged since installation, removes the managed blocks and deletes files that only contained a block. A file you edited after installation is kept and listed in the plan. Folders left empty are removed.
Uninstall does not put back the entries the installer removed from ~/.agents/.skill-lock.json. It also cannot restore a skill that JADO replaced during install: that skill exists only in the backup folder the install printed, under your system's temporary directory. Copy that folder somewhere safe after installing if you may want those skills again.
A local install copies the method into a repository, so the project carries its own procedures:
npm run jado -- install --local --target /path/to/project --agents claude-code
It writes .jado/, project-level profiles in .claude/agents/, .codex/agents/ and .opencode/agents/, and an entry in the project instructions. An existing .codex/config.toml is left unchanged; the Codex registration to merge is saved in .jado/agents/codex-registration.toml. Local installs also accept other skills CLI agent IDs, which receive skills only.
A link points a project at the global install instead of copying it:
npm run jado -- link --target /path/to/project
Use a local install when the AI tool can only read files inside the repository.
Remove either with npm run jado -- uninstall --local --target /path/to/project.
When the project already has design instructions, ask the agent to complete the setup. It follows migration.md and the jado-install skill and shows a diff before replacing older design routing.
Every command prints its plan first. For an AI-assisted install, ask for JSON, review it, then apply that exact plan:
npm run jado -- install --global --agents claude-code --dry-run --json
npm run jado -- install --global --agents claude-code --yes --revision HASH
HASH is the revision field of the reviewed plan. If a file changes between the two commands, the apply step stops. Each apply keeps a backup and restores the previous files if a write fails.
Pull the checkout first, then update the installation:
git pull
npm ci
npm run jado -- update --global
npm run jado -- update --local --target /path/to/project
--component method updates the method and keeps the installed upstream skills. --component skills updates the upstream skills and keeps the method. A global update does not change local installs.
Refresh downloads new versions of the upstream skills and leaves the method alone:
npm run jado -- refresh --global --dry-run
npm run jado -- refresh --global --yes
On macOS you can schedule it:
npm run jado -- refresh --global --enable --every-days 7
npm run jado -- refresh --global --enable --every-days 7 --check-only
npm run jado -- refresh --global --disable
--enable writes a LaunchAgent at ~/Library/LaunchAgents/com.jado.refresh.<hash>.plist. The settings and log are ~/.jado/refresh.json and ~/.jado/refresh.log. --check-only reports updates without applying them. The schedule runs from the checkout, so keep it in place.
jado/skill-lock.json lists the skills that come from other authors and the repository of each one, and the installer downloads the skills with the skills CLI during install, update and refresh. The lock does not pin commits, so each install gets the version the source repository serves at that moment. Each skill keeps its author's license. THIRD-PARTY-NOTICES.md lists the repositories and their licenses.
The skills JADO owns live in .agents/skills/ and are covered by this repository's license.
The playbook is a Next.js and Fumadocs site with the landing page at /, the docs at /docs and the workflow overview at /workflows. Each docs page is also served as Markdown at the same path with .md, and /llms.txt lists them.
npm run dev # http://127.0.0.1:3000
npm run build
npm run start
To deploy on Vercel, import the repository and keep the default Next.js settings. Set PLAYBOOK_BASE_URL to the public URL, for example https://jado.example.com. It is read at build time for canonical links, social images and the sitemap. Without it the site uses the Vercel production domain, or http://localhost:3000 outside Vercel.
Method changes are edits to the sources followed by regeneration:
jado/workflows.json: routes, steps and conditions.jado/agents.json: the five jado-agent-* profiles and the procedures they load.jado/policies/: the stage, involvement and budget settings and the protected checks..agents/skills/: the skills JADO owns.content/docs/: hand-written playbook pages.Then run:
npm run jado:compile # profiles, runtime packets, catalog, workflow pages and diagrams
npm run sync:catalog # skill and agent catalog pages
npm run check
npm run build
npm run check starts with jado:check, which fails when a generated file no longer matches its source. It then validates content and Mermaid diagrams and runs the tests, lint and type checks. Do not edit generated files by hand.
npm run jado:import adds upstream skill references to the lock. By default it reads the skills installed in ~/.agents/skills, the folder the skills CLI uses for global installs; pass --skills-root DIR to read another folder.
npm run jado:configure is optional and Codex-specific. It writes an MCP server entry to this checkout's .codex/config.toml (ignored by Git) that exposes the catalog and workflow validation as read-only tools.
| Path | Content |
|---|---|
jado/ | Workflow graph, profiles, policies, generated runtime packets and the skill lock |
.agents/skills/ | Skills JADO owns |
.codex/agents/ | Generated Codex profiles for this checkout |
registry/ | Catalog data for the site and the installer |
scripts/jado/ | CLI, installer, compiler and MCP server |
content/docs/ | Playbook pages |
app/, components/, lib/ | Site code; the landing page is in app/(landing), components/landing and public/landing |
templates/ | Reusable documents for product repositories |
MIT. Upstream skills keep their own licenses; see THIRD-PARTY-NOTICES.md.