The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python script via a bash heredoc to generate .excalidraw files. This is the primary intended function of the skill to facilitate file generation.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where it ingests untrusted user input to generate diagram labels.
Ingestion points: User-provided text descriptions and labels for diagrams (SKILL.md).
Boundary markers: None explicitly defined in the prompt instructions to delimit user content.
Capability inventory: The skill uses a Python script (scripts/generate_excalidraw.py) to write files to the local filesystem using write_scene.
Sanitization: The Python script uses the standard json.dump method for serialization, which safely handles special characters within the generated JSON structure.
[EXTERNAL_DOWNLOADS]: The documentation references an external MCP server hosted on Vercel (excalidraw-mcp-app.vercel.app) and the official Excalidraw MCP GitHub repository. These are well-known services and trusted sources used for extending the agent's diagramming capabilities.

excalidraw-diagrams