What problem does it solve? Explaining algorithms, architectures, or code changes with hand-drawn diagrams is slow and error-prone, and nothing verifies that the picture actually matches the code or the claim it makes. This Skill lets you write one declarative JSON scene and get a compiled, semantically checked animation or figure. ## Core Features & Use Cases - Scene compilation and checking: Write a JSON scene with a kind (sort, tree, state-machine, modules, sequence, and 14 more) plus intent, then run vlmkit-anim check, layout, explain, and why to validate semantics, geometry, and narration. - Code-grounded diagrams: Generate module maps, repository architecture diagrams, PR change maps, and diffs between two maps, verified against fact sheets extracted from the actual import graph via --expect. - Multiple outputs: Render still SVG/PNG figures, standalone HTML pages, GIF/MP4 videos, mermaid imports, and Markdown embeds through a remark plugin. - Use Case: Ask how a branch changes a repository; the Skill runs vlmkit-anim pr --base origin/main to produce one animated beat per commit with a paste-ready Markdown summary. ## Quick Start Ask the agent to animate how bubble sort works on the values [5, 3, 8, 1] and produce a checked SVG figure for the docs.
