The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the excalidraw-cli package globally from github:opencoredev/excalidraw-cli using the bun package manager.
[COMMAND_EXECUTION]: The skill uses various shell commands to interact with the excalidraw CLI. This includes instructions to start servers in the background using nohup and the option to bind the service to all network interfaces (0.0.0.0), which may expose the local drawing server to the network.
[DATA_EXFILTRATION]: The excalidraw url command allows the agent to upload canvas content to excalidraw.com to generate an encrypted, shareable link. While this is a documented feature for diagram sharing, it involves transmitting local data to an external service.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data through Mermaid diagrams and JSON batch files, creating a surface for indirect prompt injection where malicious instructions could be embedded in the data.
Ingestion points: The excalidraw mermaid command (reading Mermaid diagram files or stdin) and the excalidraw batch command (reading JSON diagram definitions from files or stdin).
Boundary markers: The instructions do not define specific delimiters or "ignore" instructions for the agent when processing these external data formats.
Capability inventory: The skill can execute CLI commands, read and write files, capture screenshots, and perform network uploads via the url command.
Sanitization: There is no mention of validation, escaping, or sanitization of the input diagrams before they are processed by the tool or presented to the agent.

excalidraw