By novarii
A suite of web-design extraction skills: faithfully replicate any live web component — WebGL shader backgrounds, JS-driven inline SVG, blended CSS composites — by extracting the real shipped design (shaders, generators, computed styles, live GPU/scene state) and verifying with quantitative oracles instead of eyeballing pixels.
Make a faithful replica of a component whose look is an EMERGENT COMPOSITE of CSS layers — stacked gradients, mix-blend-mode, masks, opacity, pseudo-element overlays — possibly sitting over a <canvas> or <svg>, by extracting the COMPUTED runtime styles and the full layer/stacking/blend graph, not the authored stylesheet. Use when the look comes from several coupled CSS layers blending together (the appearance is more than any single layer), when blend modes / pseudo-element overlays / masks are involved, or when a canvas/SVG looks wrong even though its own source matches. Do NOT use for a single declarative element with no blending (just copy its computed style) — this skill is for the coupling.
Make a pixel-perfect, dependency-free replica of an inline-SVG component whose geometry is regenerated by JavaScript (cursor-, scroll-, or time-driven) by reading its actual generator source from the page bundle, instead of inferring a parametric model from measured curves. Use when the thing you need to clone, extract, or replicate is an inline <svg> with real <path>/shape children whose motion is driven by JS (no <canvas>, no <animate>/SMIL). Do NOT use for <canvas> (WebGL/canvas2d), declarative CSS/SMIL keyframes, Lottie/Rive, or <img>/<video> — those are sibling extraction skills with a completely different strategy.
Verify a replicated/extracted web component against its live original with quantitative oracles instead of the lossy human eye — the SHARED verify loop for every extraction modality (webgl, css-composite, svg-js-generator, canvas2d). Decides whether a mismatch is a real defect or just animation phase, proves which layer is at fault, and turns "looks a bit off" into a measured signal. Use after you have a replica and a live original to compare, ESPECIALLY when the component animates, when single-frame screenshots disagree but you can't say why, or when a user says "the colors/brightness are a little off."
Make a faithful, deployable replica of a WebGL/WebGL2 canvas component — particle fields, point-cloud globes, shader backgrounds, GPU-driven animations — by reading its real shaders + setup source, instrumenting its live GL context to capture exact uniforms/camera/scene-graph, and reusing its downloadable assets, instead of eyeballing pixels. Use when the thing you need to clone, extract, or replicate is a <canvas> rendered with WebGL (Three.js or raw WebGL2). Do NOT use for inline SVG, declarative CSS/SMIL, Lottie/Rive, or video — those are sibling extraction skills with a completely different strategy.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A suite of web-design extraction skills. Make faithful, deployable replicas of any live web component — shader backgrounds, JS-driven SVG, blended CSS composites — by extracting the real shipped design (the GLSL, the generator source, the computed CSS, the captured GPU/scene state) and verifying with quantitative oracles, instead of eyeballing pixels and guessing a model.
Driver throughout: browser-harness (heredoc form, on $PATH); analysis via uv run.
| Command | Use it when the target is… |
|---|---|
/component-extractor:webgl | a <canvas> rendered with WebGL/WebGL2 (shaders, particle fields, globes). Read the GLSL + instrument the live GL context for real uniforms / scene graph. |
/component-extractor:svg-js-generator | an inline <svg> whose geometry is regenerated by JS (cursor/scroll/time). Read the generator source from the bundle; transcribe verbatim. |
/component-extractor:css-composite | an emergent composite of CSS layers (stacked gradients, mix-blend-mode, masks, pseudo-element overlays), possibly over a canvas/svg. Extract computed styles + the full stacking/blend/isolation graph. |
/component-extractor:verify | any replica that needs checking against a live original (esp. animated). Quantitative oracles; temporal flicker-vs-phase; isolate-and-measure. |
Components are often hybrids — extract each layer with its skill, then compose; the
difficulty lives in the seams. Each skill carries its helper code under scripts/ and
worked runs under examples/ (read on demand — progressive disclosure).
/plugin marketplace add novarii/component-extractor
/plugin install component-extractor@component-extractor
Then the four skills are available as /component-extractor:<skill> and Claude
auto-invokes them by their description.
claude --plugin-dir /path/to/component-extractor-plugin
Run /reload-plugins after editing skills.
.
├── .claude-plugin/
│ ├── marketplace.json
│ └── plugin.json
└── skills/
├── webgl/ SKILL.md scripts/{_glcap,_scenecap}.js examples/
├── svg-js-generator/ SKILL.md examples/
├── css-composite/ SKILL.md examples/
└── verify/ SKILL.md scripts/{temporal,diff}_oracle.py examples/
npx claudepluginhub novarii/component-extractor --plugin component-extractorClaude Code skills for Godot 4.x game development - GDScript patterns, interactive MCP workflows, scene design, and shaders
AI-powered development tools for code review, research, design, and workflow automation.
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
Plugin-safe Claude Code distribution of Antigravity Awesome Skills with 1,561 supported skills.