Agent skill that turns natural language, code, Terraform/K8s, SQL, OpenAPI, AsyncAPI, Protobuf and GraphQL sources into editable, tested draw.io architecture diagrams: incremental sync, multi-view projection, drift diff, CI architecture tests, whiteboard derasterize, interactive HTML/PPTX/Mermaid exports.
Tests
Tests that launch the Electron-based draw.io desktop CLI are opt-in because a binary can be installed yet unusable in a headless or macOS sandbox session. Run the deterministic unit suite normally, and enable native export tests only in a desktop-capable environment:
python3 -m unittest discover -s tests
DRAWIO_E2E=1 python3 -m unittest discover -s tests
Dependency-free regression tests for the bundled scripts (stdlib unittest
only — no pip install). They live at the repo root so they are not shipped
with the skill package.
# from the repo root
python3 -m unittest discover -s tests -v
Coverage
| Area | What's locked in |
|---|---|
shapesearch.py |
Soundex codes, index loads, known queries resolve, title-exact ranking (dynamodb → DynamoDB), empty on no match |
aiicons.py |
variant families, brand/token matching, -color preference with mono fallback (OpenAI) |
encode_drawio_url.py |
CJK + % round-trip (the encodeURIComponent-before-deflate fix), viewer vs editor URLs |
autolayout.py |
palette sourced from default.json (not the fallback), group tinting, explicit-style wins, --mono |
validate.py |
good .drawio passes (exit 0), dangling edge fails (exit 1) |
| importers | pyimports intra-project edge, pyclasses inheritance edge + no hard-coded colour |
Auto-layout is exercised through to_drawio() with synthetic positions, so the
suite needs no Graphviz or draw.io to run.