Tavotto
AI가 당신의 연구 그림을 그려줍니다
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 확인하기README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
Three skills for turning content into something you can show — a self-contained
HTML deck or dashboard, a vertical scroll deck built from a Markdown report, or
an Excalidraw architecture diagram. Packaged as a single plugin named visuals,
installable on six coding-agent harnesses.
Pick by deliverable, not by topic: .excalidraw file -> excalidraw-diagram;
Markdown in and a vertical scroll deck out -> md-to-scrolldeck; everything
else visual -> visualize.
Visual guides and worked examples (GitHub Pages)README 전문 보기 ↗Three skills for turning content into something you can show — a self-contained
HTML deck or dashboard, a vertical scroll deck built from a Markdown report, or
an Excalidraw architecture diagram. Packaged as a single plugin named visuals,
installable on six coding-agent harnesses.
| Skill | Invoke | What it does |
|---|---|---|
visualize | /visuals:visualize [<file-or-content>] | Writes one self-contained HTML file — deck, dashboard, infographic, poster, flowchart, timeline, carousel. Starts from a fixed skeleton and a design system, never from a blank page. The general-purpose default. |
md-to-scrolldeck | /visuals:md-to-scrolldeck <input.md> [--slides n] [--outline-only] | Compresses one Markdown document into a vertical scroll-snap deck (scrollytelling) for leadership review: progress bar, phase header, dot rail, arrow-key nav, print-ready. Prints its slide outline before writing anything. |
excalidraw-diagram | /visuals:excalidraw-diagram <topic-or-spec> | Generates .excalidraw JSON that argues visually rather than labelling boxes, renders it to PNG, looks at the render, and fixes it against a 23-item checklist before delivering. |
Pick by deliverable, not by topic: .excalidraw file -> excalidraw-diagram;
Markdown in and a vertical scroll deck out -> md-to-scrolldeck; everything
else visual -> visualize.
visualize — visual guide · usage example (Markdown to HTML output)md-to-scrolldeck — visual guide · usage example (Markdown to vertical scroll deck)excalidraw-diagram — visual guide · usage example (prompt to diagram)Both trees are in the repo — docs/skill-guides/ holds the
rendered skill references, and docs/skill-output/ keeps
each usage example's Markdown source beside its rendered page.
excalidraw-diagram also has its own
README covering VSCode setup and the
render pipeline.
/plugin marketplace add dEitY719/visuals-skills
/plugin install visuals@visuals-skills
codex plugin install dEitY719/visuals-skills
kimi plugin install dEitY719/visuals-skills
hermes plugins install dEitY719/visuals-skills
See .opencode/INSTALL.md.
gemini extensions install https://github.com/dEitY719/visuals-skills
Antigravity (agy) shares ~/.gemini, so it inherits the install.
npx skills add https://github.com/dEitY719/visuals-skills
These skills are written in Claude Code's vocabulary, but they lean on very
little that is Claude-Code-specific: they read files, write one output file, and
run a shell command. The per-harness tool mappings live in the sibling repo
dEitY719/harness-skills;
read the one file for the harness you are on.
| Skill | Claude Code | Codex | Kimi | Gemini / Antigravity | Hermes | OpenCode |
|---|---|---|---|---|---|---|
visualize | full | full | full | full | full | full |
md-to-scrolldeck | full | full | full | full | full | full |
excalidraw-diagram | full | needs image read-back | needs image read-back | needs image read-back | needs image read-back | needs image read-back |
needs image read-back — Step 3 renders the diagram to PNG and then looks at
it to catch overlapping text and misaligned arrows. A harness that cannot read
an image back still produces the .excalidraw and the PNG, but must report the
visual audit as skipped rather than claiming 23/23 quality items passed. That
step also needs uv and a Playwright Chromium on the machine.
Auto-open (xdg-open / open) is a no-op in a headless session; the skills
report the file:// path instead.
Manifests live at the repo root and all point at one flat skills/ directory:
.
├── skills/{visualize,md-to-scrolldeck,excalidraw-diagram}/
│ ├── SKILL.md
│ ├── references/
│ └── examples/ · evals/
├── docs/ GitHub Pages guides + samples
├── .claude-plugin/{marketplace,plugin}.json Claude Code
├── .codex-plugin/plugin.json Codex
├── .kimi-plugin/plugin.json Kimi CLI
├── .hermes-plugin/{plugin.yaml,__init__.py} Hermes Agent
├── .opencode/plugins/visuals.js + INSTALL.md OpenCode
├── .agents/plugins/marketplace.json Antigravity
├── gemini-extension.json + GEMINI.md Gemini CLI
├── package.json
├── CLAUDE.md · AGENTS.md -> CLAUDE.md
└── LICENSE
Only Claude Code understands a nested plugins/<name>/skills/ layout. The other
five harnesses resolve manifests at the repo root and a skills tree at
./skills/, so this repo keeps everything flat — the nested layout it shipped
with through v0.4.0 is gone. See CLAUDE.md for the full rationale
and contribution rules.
The .kimi-plugin/ manifest is pre-provisioned: Kimi CLI is not installed on the
maintainer's machines yet, and shipping the manifest now costs nothing and saves
a migration later.
.github/workflows/validate.yml is a uses:
call to
harness-skills/.github/workflows/skill-check.yml,
the same reusable workflow every other dEitY719/*-skills repo validates
through — manifests, skill frontmatter, progressive-disclosure line limits, the
Codex description budget, version agreement across all seven version-bearing
manifests, plugin-name consistency, the AGENTS.md symlink, the flat layout,
and shell scripts.
That shared workflow bans emoji in tracked text, and this repo's visualize
skill ships example HTML — posters, decks, infographics — that uses emoji as
intentional design glyphs in the rendered artwork. validate.yml exempts
exactly those paths with the workflow's allow-emoji-paths input rather than
inlining every other check to omit just that one. The repo-specific checks — not
part of the shared workflow — run via the tests/*.sh convention the shared
workflow auto-discovers: lib/verify-html.sh --selftest through
tests/verify-html-selftest.sh, and the
guide-versus-SKILL.md drift guard in
tests/skill-guides.sh, and
tests/vendor-drift.sh, which fails when a skill's
vendored lib/vendor/verify-html.sh differs from lib/verify-html.sh. Originally
tracked as
harness-skills#2.
These skills were extracted from
dEitY719/dotfiles
(claude/skills/devx-{visualize,md-to-scrolldeck,excalidraw-diagram}) as a
content snapshot — no history rewriting. The source commit SHA is recorded in
this repo's conversion commit message. The devx- prefix is dropped here
because the plugin namespace (visuals:) now supplies it. The dotfiles copies
stay in place until Phase 4 of that repo's migration plan.
visualize and excalidraw-diagram originate upstream — visualize by
careerhackeralex, excalidraw-diagram by
coleam00;
md-to-scrolldeck was written by @dEitY719.
The manifests carry only author: dEitY719 (the packager) and license: MIT,
because Claude Code's manifest schema rejects any other top-level field, so
this section is where upstream authorship is recorded.
This is Phase 1 of the dotfiles #1410 migration; packaging-skills was Phase 0
and harness-skills is its sibling in this phase.
MIT. See LICENSE.