// Extract a real design system from a live URL into `design.md` and `report.ko.html`, or reuse the bundled multi-site design corpus as a deterministic design brief. Use when the user wants real CSS tokens, a reusable design reference, or a design pack that can feed apply/build flows.
| name | insane-design |
| description | Extract a real design system from a live URL into `design.md` and `report.ko.html`, or reuse the bundled multi-site design corpus as a deterministic design brief. Use when the user wants real CSS tokens, a reusable design reference, or a design pack that can feed apply/build flows. |
Read these first:
shared/README.mdreferences/workflow.mdreferences/pitfalls.mdreferences/apply-mode.mdreferences/build-mode.mdLoad before writing output:
references/template.md for design.mdreferences/report-prompt.md and references/report.css for report.ko.htmlUse examples/ as the canonical bundled corpus. Do not duplicate or prefer any published mirror.
analysis — extract a fresh design.md and report.ko.html from a live URLapply — apply an existing design.md or bundled example to an existing projectbuild — generate a new artifact from a design.md plus starter componentsapply and build are internal modes of insane-design, not separate top-level skills.
insane-design/<slug>/.scripts/.references/data-collection.md; do not hallucinate tokens you can extract.scripts/brand_candidates.pyscripts/var_resolver.pyscripts/typo_extractor.pyscripts/alias_layer.pydesign.md with the exact v3.1 frontmatter and §18 DON'T contract from shared/README.md.report.ko.html with the canonical report prompt and CSS.phase1 artifacts when they support later apply/build work.apply and build modes depend on the frontmatter and §18 DON'T contract staying stable.Structured research-query builder for turning a vague topic into a research brief and machine-readable query before full deep research starts. Use when the user wants help framing a research request.
Comprehensive, citation-heavy research workflow with session state, structured outputs, and source-quality checks. Use when the user explicitly asks for deep research, a long-form report, or citation-backed analysis on a topic.
Fetch and explain official documentation for libraries, frameworks, and tools using llms.txt-first retrieval and official-source fallbacks. Use when the user wants official docs, version-aware references, or an explanation grounded in primary documentation.
Explain Git and GitHub concepts using cloud-folder analogies for non-developers. Use when the user asks "commit이 뭐야?", "push랑 commit 차이", "PR이 뭐야?", "branch가 뭐야?", "깃 용어", "도움말", or similar beginner Git questions.
Create a beginner-friendly GitHub Pull Request workflow. Use when the user says "PR 만들어줘", "검토 요청", "리뷰 요청", "팀원한테 보여주고 싶어", "create PR", or "pull request".
Stage and commit changed files with beginner-friendly Korean explanations. Use when the user says "저장해줘", "커밋", "변경 내용 저장", "save", or "commit".