← ClaudeAtlas

excalidraw-diagramslisted

Author an Excalidraw diagram for an Atlas Visuals digest — when a note explains a flow/sequence/architecture, compose a scene with the Excalidraw MCP, save it under {visualsDir}/files/diagrams/, and embed it with <Diagram src>.
muslewski/memory-atlas · ★ 3 · AI & Automation · score 69
Install: claude install-skill muslewski/memory-atlas
# Excalidraw Diagrams for Atlas Visuals The `<Diagram>` kit primitive embeds an Excalidraw scene read-only inside an Atlas Visuals digest (package **memory-atlas-visuals**). Use this skill when authoring the scene itself. Resolve paths from the consuming repo: | Token | Source | |-------|--------| | `{vaultDir}` | `atlas.config.json` → vault dir | | `{visualsDir}` | `config.visuals.dir` if set, else `{vaultDir}/visuals` | | `{appDir}` | `{visualsDir}/app` | CLI gates: `pnpm exec atlas-visuals …` or `npx atlas-visuals …` (or `pnpm --dir {appDir} …`). ## When to reach for a diagram **Default-on: every digest gets a diagram of its core structure.** A picture makes a human grasp the note at a glance — that is the whole point of the visual layer. Pick the angle that most *clarifies* the note, not always a flow: - **flow / pipeline / sequence** — steps that feed each other (render loop, request path) - **relationship map** — how parts connect (zones, substrates, ownership) - **hierarchy / tree** — nesting or containment (collections, layers) - **lifecycle / state machine** — a thing moving through states (content, jobs) - **decision / branch** — a fork and its paths (manual vs delegate) - **before → after** — a failure and its fix (the cascade-delete diagram) **One diagram = one idea**, and it must *clarify* (not decorate) — prose still carries the nuance, the diagram carries the shape. Skip a diagram ONLY when a note genuinely has no structure to draw (a flat list with no