The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads necessary libraries from the public npm registry.
Evidence: scripts/setup.sh executes npm install to fetch @excalidraw/excalidraw, jsdom, and canvas.
[COMMAND_EXECUTION]: The skill executes local scripts to process diagram files.
Evidence: User instructions involve running node scripts/render.mjs and node scripts/lint-scene.mjs to generate and validate scenes.
[PROMPT_INJECTION]: The skill ingests untrusted data that could contain indirect instructions.
Ingestion points: Reading element text from .excalidraw JSON files.
Capability inventory: The agent uses this data to summarize or reason about diagrams, and the toolchain performs subprocess calls via node for rendering.
Boundary markers: None identified for text element content.
Sanitization: JSON parsing is performed, but no semantic filtering or escaping of text content is noted for agent interpretation.

excalidraw