What problem does it solve? Hand-authored architecture and methods figures in matplotlib repeatedly hit the same visual problems: overflowing box labels, inconsistent palettes, spindly arrows, broken SVG exports in Illustrator, and ad-hoc figure sizes. This Skill encodes hard-won layout and styling lessons as reusable primitives so every figure starts from correct defaults. ## Core Features & Use Cases - Diagram primitives: stage_box, dual_stage_box, stack_box, card, oval, circle, labeled_arrow, lane_arrow, perf_gauge, legend_block, and fit_text for auto-shrinking overflowing labels. - Gantt and lifecycle support: gantt_bar, today_line, variance overlays, two-tier snapshot diffs, fork-join parallel-track routing, and graphviz-driven autolayout via dot. - Three palettes: flat ACCENT actor palette, modern CARD_* card palette with drop shadows, and colour-blind-safe Okabe-Ito / Tol muted scientific palettes on a warm-paper base. - SVG round-trip hygiene: Helvetica-only fonts, ASCII-safe glyphs, svg_safe() clip stripping, and editable text for clean Illustrator imports. - Use Case: Ask for a side-by-side pipeline architecture figure contrasting LLM and deterministic stages with a performance gauge, and get a 16:9 SVG that opens cleanly in Illustrator. ## Quick Start Use the architecture-figures skill to draw a left-to-right pipeline diagram with three stages, an LLM stage in blue, a deterministic stage in slate, and a labeled feedback arrow, exported as SVG.