Back to search
ryugi-dev
MarketplaceWorks with Claude CodeGitHub
Development marketplace for the Ryugi skills and artifacts workbench
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/innei--ryugi)HTML
<a href="/listings/innei--ryugi"><img src="/api/listings/innei--ryugi/badge.svg" alt="findsafeskills safety rating" /></a>Install
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 31 lines
- {
- "description": "Development marketplace for the Ryugi skills and artifacts workbench",
- "name": "ryugi-dev",
- "owner": {
- "name": "Innei",
- "email": "i@innei.dev"
- },
- "plugins": [
- {
- "name": "ryugi",
- "description": "Ryugi skills for coding agents, plus the artifacts workbench they design in: brainstorming, artboards, documents and TDD.",
- "version": "1.85.0",
- "source": "./",
- "author": {
- "name": "Innei",
- "email": "i@innei.dev"
- }
- },
- {
- "name": "ryugi-artifacts",
- "description": "Ryugi artifacts only: artboards of mockups, documents and HTML pages you comment on, with comments delivered back to the agent. No workflow skills.",
- "version": "1.85.0",
- "source": "./",
- "author": {
- "name": "Innei",
- "email": "i@innei.dev"
- },
- "skills": ["./skills/artifacts", "./skills/interface-craft"]
- }
- ]
- }
README.md · README · 243 lines
- <p align="center">
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="assets/logo/ryugi-wordmark-dark.svg">
- <source media="(prefers-color-scheme: light)" srcset="assets/logo/ryugi-wordmark-light.svg">
- <img src="assets/logo/ryugi-wordmark-light.svg" width="240" alt="Ryugi">
- </picture>
- </p>
- <h1 align="center">Ryugi 流儀</h1>
- <p align="center"><i>ryoo-ghee</i> · [ɾʲɯːɡʲi] · Japanese for “a school's own way of doing things”</p>
- <p align="center"><b>Artifacts for coding agents.</b><br>Your agent draws its options instead of describing them. You pick, comment and pass.</p>
- <p align="center">
- <img src="docs/images/hero.webp" alt="A terminal asks Claude Code to mock up notification settings; three layouts stream onto the Ryugi artboard, the user picks the recommended one and submits, the terminal receives q1 → b, and the built page appears next to the approved design, where the user presses Pass">
- </p>
- Ask an agent for a settings page and you get three paragraphs describing three layouts. Ask for a
- design and you get a wall of Markdown you can't point at.
- Ryugi gives the agent pages to put that work on, and gives you a way to answer on them:
- - **Artboard** — every option drawn side by side; pick one.
- - **Document** — a design doc with the agent's questions inline; comment on any sentence.
- - **Whiteboard** — boxes and arrows you both draw on.
- - **HTML** — one page with real sliders to tune a component or a motion.
- - **Slides** — a deck the agent writes and you present, with your notes on a second screen; fix
- any slide by hand.
- - **Verify** — the built screen next to what you approved, and a report of what the agent proved
- by running it; pass or reject.
- Your picks, comments and verdicts go straight back to the agent session that made the page. Every
- artifact lives in your project, opens from one local server, and becomes a link when you want to
- share it.
- ```bash
- npx ryugi-artifacts install
- ```
- The terminal wizard installs a persistent `ryugi` command, lets you select one or more
- harnesses, then asks which content to install:
- - **Artifacts only** — send `Show me three directions for a settings page` and an artboard opens.
- - **The full workflow** — artifacts plus skills that brainstorm, spec, build under TDD and verify.
- Send `Let's make a react todo list` and the agent opens an artboard before it writes code.
- <p align="center">
- <a href="docs/install.md">Harnesses</a> · <a href="docs/artifacts-cli.md">CLI</a> · <a href="docs/workflow.md">Full workflow</a> · <a href="https://ryugi.innei.dev">Ryugi Cloud</a>
- </p>
- ## Artboard — choose instead of imagine
- **Use it when** there is more than one way it could look.
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="docs/images/artboard-dark.webp">
- <img src="docs/images/artboard-light.webp" alt="A choice group of three notification settings layouts on the artboard, the per-event grid marked Recommended and chosen, a sticky note with the reason, and the round bar reading 1/1 chosen with a Submit button">
- </picture>
- The agent draws every option side by side and marks the one it recommends. Pick, comment on any
- pixel, and Submit sends your answer back.
- ## Document — design docs you can argue with
- **Use it when** the decision is structure, architecture or tradeoffs.
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="docs/images/document-dark.webp">
- <img src="docs/images/document-light.webp" alt="A design document for notification settings: an inline question answered with “a daily digest”, a link to the chosen board, and a margin comment from Claude Code marked Waiting on you">
- </picture>
- The agent asks its questions inside the doc and links the boards it drew. Comment on a sentence or
- edit in place; your edits and the agent's merge live.
- ## Whiteboard — sketch the structure together
- **Use it when** the parts and arrows are still the question.
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="docs/images/whiteboard-dark.webp">
- <img src="docs/images/whiteboard-light.webp" alt="A whiteboard with the agent's delivery flow — Event, Queue, then Email digest, Push and In-app — and a box the user drew asking “Rate limiter?” with the agent's answer beside it">
- </picture>
- You both draw on it. Select what you drew, ask about it, and the answer lands next to your sketch.
- ## HTML — tune it with real controls
- **Use it when** one thing needs tuning: a component, a type scale, a motion.
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="docs/images/html-dark.webp">
- <img src="docs/images/html-light.webp" alt="A test bench for the toggle switch spring: stiffness, damping and mass inputs, two toggles and a plotted settle curve with settle time and overshoot">
- </picture>
- A one-page bench whose sliders keep their values when the agent rewrites the page.
- ## Slides — a deck you can present
- **Use it when** you will stand in front of people with it: a talk, a team share-out, a launch.
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="docs/images/slides-dark.webp">
- <img src="docs/images/slides-light.webp" alt="A six-slide deck on notification settings: the statement slide “A daily digest by default, instant push only for what can't wait” on stage, the filmstrip below, and Present and Presenter mode in the header">
- </picture>
- The agent writes the deck and it follows your light or dark look. Present it full screen, or in
- presenter mode with your notes and the next slide on your screen and the audience on another. Click
- a slide to fix a word, restyle or move a block yourself, or pin a comment for the agent.
- ## Verify — you pass or reject
- **Use it when** the work is built and you decide whether it matches.
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="docs/images/verify-dark.webp">
- <img src="docs/images/verify-light.webp" alt="A verify group with the approved per-event grid next to the built settings page, and Pass and Reject on the group label">
- </picture>
- The built page sits next to the design you approved. Pass it, or reject it and say what is off; it
- comes back fixed.
- What a screen can't show goes in a **verify report**: one line per claim the agent checked by
- running something — a test run, an API response, a migration count, which build is deployed — with
- the raw output one click away, and a "Not verified" list of what it couldn't check and why. You read
- it top to bottom and pass, reject or waive; screens still get judged on the artboard, and the
- report links to them.
- ## Comments go back to the agent
- **Use it when** you'd rather point at the thing than describe it in the terminal.
- Pin a comment on a board, a sentence or a sketch, and it reaches the agent session that opened the
- page. In Claude Code it arrives on its own, with no message from you; the agent applies it, replies
- under your comment and updates the page while you watch. Questions, picks and verdicts travel the
- same way.
- ## Library — every artifact, one server
- **Use it when** you want last week's mockup back, or a link to share.
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="docs/images/library-dark.webp">
- <img src="docs/images/library-light.webp" alt="The artifact library for one project: type filters, search, and cards for an artboard, a document, a whiteboard and an HTML bench">
- </picture>
- ```bash
- npx ryugi-artifacts
- ```
- Local first, no subscription, no agent session needed. One server covers every project on the
- machine, with ⌘K to jump and links that survive a restart. `publish` shares one to
- [Ryugi Cloud](https://ryugi.innei.dev).
- For a persistent `ryugi` command without a harness plugin:
- ```bash
- npx ryugi-artifacts install --mode artifacts --harness none
- ryugi
- ryugi /path/to/project
- ```
- Node.js 20.12 or newer is required. The installer keeps a complete runtime outside
- the npx cache and creates `~/.local/bin/ryugi` (a command shim on Windows). It prints
- PATH instructions for your shell. `npm install -g ryugi-artifacts` remains available;
- package-manager installations keep their original update channel. `ryugi update`
- updates the CLI and recorded plugins; `ryugi doctor` checks the install and
- `ryugi uninstall` removes owned entries while retaining artifacts and user data.
- See the [CLI guide](docs/artifacts-cli.md) for commands and options.
- ## Add the full workflow
- Artifacts work on their own. If you also want the agent to build from what you approved, pick the
- full workflow at install (or run `ryugi install --mode full --harness codex,pi` later). Its skills make
- the agent reach for the artifacts at the right moment, and they fire on their own:
- ```mermaid
- flowchart LR
- I[Idea] --> B[Brainstorm] --> A[Artifacts] --> S[Spec] --> T[Build · TDD] --> V[Verify] --> L[Land]
- ```
- After you sign off, the agent slices the spec into tasks, proves each one red-green, keeps a ledger
- of judgment calls across compaction, gets one fresh review of the branch, and reports how to land
- it. Debugging and worktrees are covered too. [How it works →](docs/workflow.md)
- ## Work with your harness
- Have Node.js and npm available in the agent's environment. The plugin calls its own launcher,
- which downloads `ryugi-artifacts` on first use and caches each release under
- `~/.cache/ryugi/artifacts/` by default. It checks for updates at most once a day and can reuse a
- compatible cached release offline. The first run needs network access and a writable cache.
- ### Claude Code settings
- Ryugi replaces Claude's hosted Artifacts with the Artboard, documents, whiteboards, HTML
- benches and decks. Turn the hosted ones off by adding this key to the top level of `~/.claude/settings.json` —
- keep your other settings:
- ```json
- {
- "disableArtifact": true
- }
- ```
- That is the only flag Ryugi needs; leave the rest of your Claude Code settings alone.
- <details>
- <summary>Other harnesses</summary>
- | Harness | Command |
- | ------------- | ----------------------------------------------------------------------------------------------------- |
- | Antigravity | `agy plugin install https://github.com/Innei/ryugi` |
- | Codex, Cursor | Add this repo as a local plugin source ([detail](docs/install.md#codex-cursor)) |
- | Devin CLI | `devin plugins install Innei/ryugi` |
- | Factory Droid | `droid plugin marketplace add https://github.com/Innei/ryugi` then `droid plugin install ryugi@ryugi` |
- | Gemini CLI | `gemini extensions install https://github.com/Innei/ryugi` |
- | Kimi Code | `/plugins install https://github.com/Innei/ryugi` |
- | OpenCode | [.opencode/INSTALL.md](.opencode/INSTALL.md) |
- | Pi | `pi install git:github.com/Innei/ryugi` |
- | Qwen Code | `qwen extensions install Innei/ryugi` |
- | Hermes Agent | `hermes plugins install Innei/ryugi --enable` |
- | Muse | `muse plugins install ./` then `muse plugins approve ryugi` |
- Updates and caveats: [docs/install.md](docs/install.md).
- </details>
- ## Credits
- Fork of [Superpowers](https://github.com/obra/superpowers) v7.0.0 by Jesse Vincent and [Prime Radiant](https://primeradiant.com) ([UPSTREAM.md](UPSTREAM.md)). [Issues](https://github.com/Innei/ryugi/issues) and PRs welcome; [AGENTS.md](AGENTS.md) has the rules.
- ## License
- Skills, harness integrations and docs: [Ryugi License 1.0](LICENSE) — free to
- use, modify and redistribute, with two exceptions. The Restricted Parties in
- Schedule A (Tencent, Alibaba, ByteDance, Baidu, Huawei, Xiaomi, Meituan, JD,
- PDD, NetEase, Kuaishou, Didi, Ant Group, iFlytek, SenseTime, and products such
- as WorkBuddy, Qwen, Qoder and Trae) receive no rights at all, and nobody may
- scrape the skills into a marketplace. This is source-available, not an open
- source license; material derived from Superpowers and contributions made before
- 2026-11-04 stay MIT (Exhibit B).
- The artifacts application, published to npm as [`ryugi-artifacts`](https://www.npmjs.com/package/ryugi-artifacts), is proprietary (npm `license`: `UNLICENSED`) and not part of this repository. Releases up to 1.51.2 were published under AGPL-3.0-only.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.