What problem does it solve? Answering "how does this work" questions about code with prose alone often fails when the answer involves many modules, states, or steps and their relations. This Skill produces a drawn figure (module map, sequence walk, state machine, algorithm trace) generated from the actual code by vlmkit-anim, verified against the facts it depicts, and paired with prose that walks the figure beat by beat. ## Core Features & Use Cases - Fact-grounded diagrams: Extracts the real import graph via vlmkit-anim facts or repo, then validates the scene with check --expect so the picture matches the code rather than memory. - Multiple scene kinds: Module maps, sequence walks, state machines, flowcharts, distributed-system messages, Gantt charts, algorithm visualizations, PR change maps, and Mermaid imports. - Checked delivery pipeline: Scene JSON is validated with check, layout, and explain, then rendered as an SVG still for structures or a GIF for walks, with prose that names elements by their ids. - Use Case: A user asks "how does a request flow through this service?" The Skill reads the code path, builds a modules scene with a sequence, verifies it against the import graph, renders a GIF, and delivers a beat-by-beat explanation citing the actual files and functions. ## Quick Start Ask the agent to explain how this repository is structured or how a request flows through it, and request a checked diagram with a walkthrough.
