Create and edit editable Excalidraw diagrams in Canvas Notebook through the Scene-Service tools. Use for diagrams, flows, sketches, architecture maps, visual explanations, or requested .excalidraw documents.
The English summary is being prepared. Showing the original description.
Review the instructions the agent will receive before you install.
Create and edit a managed .excalidraw document through Canvas's Scene-Service. PostgreSQL holds the authoritative scene; the portable file is a checkpoint. These tools work without an open editor in personal, organization, team, and project workspaces.
Use create_excalidraw_scene, get_excalidraw_scene, and edit_excalidraw_scene. If a required tool is missing or disabled, report an activation or compatibility error and ask the user to enable it or use a compatible Canvas backend. A missing PostgreSQL provider is an error. Do not substitute another write path.
Never create or change .excalidraw bytes through generic write/edit/patch tools, direct workspace filesystem access, shell, Python, Node, copy-overwrite, or snapshot saves. Never submit a complete scene replacement. The bundled Python validator only reads JSON; it does not create files or publish a scene.
Internal tools default to the current agent workspace. An explicit workspaceId must identify that same execution context; do not switch to another workspace by passing its ID. External MCP calls require an explicitly authorized workspaceId, enabled tools, and knowledge:read for Get or knowledge:write for Create/Edit. Canvas supplies actor identity; never send actor or permission fields.
.excalidraw path. Use a descriptive filename when the user only specifies a folder. Inspect relevant source material for technical diagrams; translate Mermaid nodes and edges into editable Excalidraw elements.path and a stable, explicit idempotencyKey. Create never overwrites an existing file. For an existing document, call Get with documentId or path; if both are supplied, they must identify the same current document.documentId, lifecycleGeneration, sceneSequence, current path, elements, and assets. Plan the layout and construct valid complete elements. Preserve existing element IDs and unknown element fields when editing.lifecycleGeneration, the returned sceneSequence as observedSceneSequence, a stable idempotencyKey, and element actions. Optional path must still match the document. Use at most 500 distinct targets and keep the request within 1 MiB. Put related elements and bindings in one batch so the resulting scene has valid references.operationId, actual status, sceneSequence, and checkpoint.status. Only applied confirms the scene changes were saved. A checkpoint can separately be saved, pending, or failed; applied with a pending/failed checkpoint still means the scene is saved in PostgreSQL. Report the checkpoint condition and do not create duplicate elements to compensate.needs_review, the entire batch remains unapplied. Read the current scene again, reassess the requested changes against current elements, then send updated actions with current versions and a new idempotency key. Never describe the proposal as applied. The experimental Review Center is disabled by default. Human Accept/Reject is an optional alternative only when the actual response says reviewAvailable: true; do not require or enable that setting as part of ordinary diagram work.documentUrl from the result with a concise outcome. Do not create a sidecar Markdown, PNG, SVG, or README unless requested. If browser verification is authorized, visually inspect the document for clipped labels and incorrect relationships.Stop on permission, identity, lifecycle, provider, or validation errors and report the actual error. A reused old path does not authorize edits to a new document identity. Do not loop on repeated target conflicts or silently discard intervening work.
Create uses a complete new element with an ID not used by any live or deleted element:
{"type":"create","element":{"id":"new_shape","type":"rectangle","version":1,"versionNonce":105,"isDeleted":false,"x":0,"y":0,"width":220,"height":100,"angle":0,"strokeColor":"#1f2937","backgroundColor":"#e0f2fe","fillStyle":"solid","strokeWidth":2,"strokeStyle":"solid","roughness":1,"opacity":100,"seed":101,"groupIds":[],"boundElements":null,"link":null,"locked":false}}
Update copies the whole read element, changes intended fields, and uses its exact observed version and nonce. The element.id must equal elementId; preserve the read version/nonce in the proposed element and let Canvas assign the committed values:
{"type":"update","elementId":"existing_shape","expectedVersion":4,"expectedVersionNonce":209,"element":{"id":"existing_shape","type":"rectangle","version":4,"versionNonce":209,"isDeleted":false,"x":20,"y":40,"width":220,"height":100,"angle":0,"strokeColor":"#1f2937","backgroundColor":"#dcfce7","fillStyle":"solid","strokeWidth":2,"strokeStyle":"solid","roughness":1,"opacity":100,"seed":101,"groupIds":[],"boundElements":null,"link":null,"locked":false}}
Create/Update elements must be live (isDeleted: false). Delete uses the exact observed version and nonce:
{"type":"delete","elementId":"existing_shape","expectedVersion":4,"expectedVersionNonce":209}
The Edit envelope is:
{"documentId":"FROM_CREATE_OR_GET","lifecycleGeneration":1,"observedSceneSequence":0,"idempotencyKey":"UNIQUE_LOGICAL_OPERATION_KEY","actions":[]}
Replace the example identity/version values with the actual response and supply a nonempty action list. Use observedSceneSequence, not a write field named sceneSequence. Do not target an element twice in one operation. Edit accepts element actions only, with no root elements, appState, files, or assets replacement fields.
[A-Za-z0-9_-]{1,128}. Use meaningful IDs with a unique suffix for newly created elements.version >= 1, integer versionNonce >= 0, boolean isDeleted, finite geometry, and complete rendering fields as in the rectangle example.text, originalText, fontSize, fontFamily, textAlign, verticalAlign, and lineHeight. Use font 5 for Excalifont, 2 for Helvetica, or 3 for code. Size boxes to fit the actual text.[x,y] points. Arrows need arrowheads. containerId, frameId, boundElements, startBinding, and endBinding must refer to elements in the resulting scene. Bound text and its container must reference each other.fileId only when required. Do not invent asset IDs, base64 files, or unfinished pending image placeholders.For a read-only preflight, pass {"scene": GET_RESULT, "request": EDIT_INPUT} as JSON on stdin to the bundled validator:
python3 <skill-directory>/scripts/validate_excalidraw.py --operation-stdin
No temporary or workspace file is needed. --stdin checks portable scene JSON from stdin; the optional positional path reads an existing portable file without changing it. These portable modes are inspection aids, not an authoring or publishing path.
Use layout to express meaning: timelines for sequences, fan-out for one-to-many, convergence for aggregation, nested regions for ownership, and side-by-side comparisons. Keep labels short, leave generous margins, and use arrows only for meaningful relationships. Include concrete endpoint names, events, filenames, or small data examples when they clarify a technical diagram.
Use color for meaning. Suggested palette: white background, #1f2937 primary stroke, #475569 secondary stroke, #e0f2fe primary fill, #dcfce7 success, #fef3c7 warning, #fee2e2 error, and #f8fafc neutral. Keep the scene within a coherent bounding box and avoid uniform card grids unless the objects are peers.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
The English summary is being prepared. Showing the original description.
Applies Canvas brand colors, typography, and visual polish to artifacts that benefit from a consistent product look-and-feel. Use it when brand colors, style guidelines, visual formatting, or company design standards apply.
The English summary is being prepared. Showing the original description.
Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.
The English summary is being prepared. Showing the original description.
Create or improve an organization-managed Canvas plugin that combines complete new or existing skills with optional custom MCP and Composio integrations. Use for reusable multi-skill workflows, plugin composition, plugin editing, and plugin graphics.
The English summary is being prepared. Showing the original description.
Use this Canvas-owned clean-room skill to guide collaborative writing for documentation, proposals, technical specs, decision records, PRDs, RFCs, operating guides, and structured memos. Trigger when the user wants help drafting, organizing, refining, reviewing, or testing a document before sharing it.
The English summary is being prepared. Showing the original description.
Use this Canvas-owned clean-room skill for creating, reading, editing, reviewing, redlining, commenting, or repairing Word `.docx` documents. Trigger for Word documents, DOCX files, memos, reports, letters, templates, tracked changes, comments, table of contents, headers, footers, page numbers, and Google Docs-targeted documents that should be authored as DOCX first.
The English summary is being prepared. Showing the original description.