The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides several Python scripts (add-arrow.py, add-icon-to-diagram.py, split-excalidraw-library.py) that are used to programmatically edit Excalidraw JSON files. The agent is instructed to execute these scripts via shell commands to perform legitimate diagram construction tasks.
[EXTERNAL_DOWNLOADS]: The skill documentation directs the user and agent to visit https://libraries.excalidraw.com/ to download official and community-provided icon sets. This is a well-known service associated with the primary purpose of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill processes external .excalidrawlib files provided by the user. If these files contain adversarial strings in element names or labels, there is a low risk of influencing the agent's behavior during the analysis of generated reference files.
Ingestion points: The script split-excalidraw-library.py reads external JSON files to generate icon metadata and a reference.md table.
Boundary markers: No specific delimiters are mandated for the agent when reading these generated reference documents.
Capability inventory: The skill has the capability to modify local files and execute shell commands via its bundled scripts.
Sanitization: Filename generation in the splitter script includes a sanitization routine that removes non-alphanumeric characters, effectively preventing directory traversal attacks via malicious icon names.

excalidraw-diagram-generator