What problem does it solve? Turning a written description of a system into a clean, readable architecture diagram normally requires manual layout work in a diagramming tool, and naive LLM generation produces cluttered, unbalanced graphs. This Skill generates a validated layered architecture diagram directly from free text, enforcing visual-quality budgets so the result stays readable. ## Core Features & Use Cases - Text-to-diagram pipeline: Generates a canonical layered graph JSON (title, layers, nodes, edges) from a free-text system description via a structured-outputs LLM call, then transforms it into graph JSON, self-contained SVG, or PNG (SVG plus export instructions). - Two-stage validation: Stage 1 enforces structural rules (3–5 layers, 6–12 nodes, edge budget of floor(n × 0.8), ID and layer-order integrity); Stage 2 checks format-level output such as SVG well-formedness and layout sanity. - Opt-in versioning: Persists diagram revisions as YAML files with a history log, supporting save, diff, rollback, and natural-language edits across long sessions. - Use Case: Paste a paragraph describing your SaaS platform (web app, API gateway, services, database) and receive a balanced 4-layer SVG architecture diagram with color-coded layer bands, ready to embed in documentation. ## Quick Start Ask the assistant to draw an architecture diagram of your system, for example: "Draw a layered architecture diagram of my e-commerce backend as SVG."
