OverviewHistoryStatsSecurity
npx skills add ...
Documentation
SKILL.md
npx skills add agents365-ai/365-skills --skill drawio-skill
Create, edit, synchronize, inspect, test, and publish editable draw.io diagrams. Use when the user explicitly requests draw.io/diagrams.net, needs a polished architecture, ERD, UML, sequence, C4, SysML, BPMN, network, swimlane, ML, or infrastructure diagram, wants code/IaC/SQL/OpenAPI/AsyncAPI/Protobuf converted into a diagram, or wants an existing diagram queried, reviewed, diffed, restyled, kept in sync, or made interactive. Prefer Mermaid/PlantUML elsewhere when the requested artifact is diagrams-as-code rather than an editable draw.io file.
npx skills add agents365-ai/365-skills --skill drawio-skill
Produce editable .drawio artifacts, not flattened pictures. The preferred
entrypoint is scripts/diagramctl.py, which unifies generation, incremental
sync, multi-view projection, semantic queries/tests/reviews, failure analysis,
and accessible publishing over a shared Diagram IR.
| Request | Route |
|---|---|
| Natural-language diagram with precise styling | Read references/diagram-types.md, then references/xml-authoring.md and author XML |
| Standard flowchart/mindmap/gantt/timeline/etc. with no special styling | If draw.io >=30, read references/mermaid-authoring.md and convert Mermaid to native .drawio |
| Large graph (~15+ nodes) that needs automatic layout | Use autolayout.py; read references/autolayout.md before passing any --layout value |
| Code, Terraform, K8s, compose, SQL, OpenAPI, AsyncAPI, or CI source | Use diagramctl.py build; read references/diagram-ir.md |
| Protocol Buffers schema (.proto) | Use protoimports.py or diagramctl.py build; read references/toolbox.md |
| Running cluster/stack/cloud (actual state, not declared config) | Read references/live-infra.md, then use tfstate.py, dockerimports.py, or k8simports.py - |
| Update a generated diagram without losing manual layout | Use diagramctl.py sync; read references/diagram-ir.md |
| Executive/system/deployment/data-flow/security views | Use diagramctl.py views; read references/diagram-ir.md |
| Query, architecture policy, review, what-if, or guided walkthrough | Read references/semantic-workflows.md |
| MCP host (Claude Desktop, Cursor, VS Code, Codex) should call these workflows | Register scripts/diagramctl_mcp.py; read references/mcp.md |
| Prompt phrasing for a diagram type or semantic workflow | Read references/cookbook.md |
| Enforce architecture rules or visual diffs in GitHub Actions CI | Read references/ci-gate.md |
| Rendered before/after/diff images as a PR review comment | Use prdiff.py; read references/pr-bot.md |
Existing .drawio to HTML/PPTX/Mermaid/Markdown/animation/runbook | Read references/toolbox.md; diagramctl.py transform exposes the existing tools |
| Pipeline, journey, or subsystem map drawn as a metro/subway map | Use tubemap.py; read references/tubemap.md |
| Shape, cloud/vendor, AI, or Databricks icon | Read references/shapes.md or references/databricks.md; never guess shape names |
| Learn/apply/manage a visual style | Read references/style-presets.md |
| Extract a reusable style from an existing diagram or theme | Read references/style-extraction.md |
| Existing image to editable diagram (screenshot, whiteboard photo, legacy PNG) | Read references/derasterize.md |
| Export/platform problem | Read references/troubleshooting.md; for access/network questions read references/security.md |
Run from this skill directory, or replace scripts/ with the absolute path to
this skill's scripts directory:
doctor does not launch GUI tools unless --probe is passed. Core semantic
commands are offline and stdlib-only.
Infer the diagram type, audience, scope, output format, and location from the
request. Ask only when a missing choice materially changes the result;
default to PNG plus .drawio in the working directory.
Select the authoring route from the table above. For a data-backed diagram,
prefer Diagram IR and preserve provenance. For a large graph, use an importer
or autolayout.py; do not hand-place more than roughly fifteen nodes.
Resolve an explicitly named style preset, or the user's default preset, as
documented in references/style-presets.md. Structural diagram conventions
and visual presets compose; they do not replace each other.
Generate the .drawio, then run structural validation:
When semantic metadata or an architecture policy is in scope, also run
diagramctl.py test. Do not present inferred semantic findings as verified
runtime facts.
Export a draft PNG without embedded XML and inspect it visually. Fix obvious overlap, clipping, disconnected edges, edge-through-node routing, stacked edges, and unreadable labels. Stop automatic vision repair after two rounds.
Show the draft and apply targeted edits. Preserve existing geometry for
local changes. Use sync for source-backed changes and write a reviewable
output; use --prune only when deletion was requested.
After approval, create final requested formats and report both editable source and export paths.
Resolve the available binary once (drawio, draw.io, the macOS app path, or
the Windows executable) and use that exact binary for the run.
Do not combine --width and -s. Embedded PNG exports require
repair_png.py; draft PNGs used by vision must not use -e. On Linux headless,
follow references/troubleshooting.md rather than improvising Electron flags.
If the CLI crashes in a macOS sandbox, try one permitted escalated run, then use
encode_drawio_url.py or deliver XML; do not repeatedly launch it.
0 or 1.<mxGeometry relative="1" as="geometry"/>.data-model-id, semantic properties, manual geometry, and
manual styles intact unless the user requests otherwise.edgeports.py; add waypoints when an
edge still crosses an unrelated shape. There is no CLI-only edge rerouter
that preserves node positions.An attractive diagram can still be wrong. Prefer source-backed relationships, show provenance where useful, distinguish exact extraction from AI inference, and keep architecture review findings framed as prompts. Story HTML must remain self-contained, keyboard usable, and include a text alternative. Never include secrets in node properties or provenance because they are embedded in outputs.
For all focused scripts and composition patterns, read references/toolbox.md;
load only the task-specific reference needed for the current request.