The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses scripts/drawio-start.sh and scripts/drawio-export.sh to execute shell commands, manage system processes via pgrep, and launch the draw.io desktop binary with specific command-line arguments.
[DYNAMIC_EXECUTION]: The core functionality in scripts/drawio-eval.mjs involves evaluating arbitrary JavaScript strings inside the draw.io Electron application's renderer process using the Chrome DevTools Protocol (CDP). This allows the agent to execute code at runtime to manipulate the diagram editor.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from diagram files and labels which could contain malicious instructions designed to influence the agent's behavior during inspection tasks.
Ingestion points: The agent reads diagram content through D.dump(), D.find(), and D.info() in scripts/drawio-helpers.js and scripts/drawio-eval.mjs.
Boundary markers: There are no explicit instructions or mechanisms to treat diagram labels as untrusted data or to ignore instructions embedded within them.
Capability inventory: The skill has the capability to write to the filesystem (D.save()), execute scripts within draw.io, and perform network requests to the local debugging port.
Sanitization: No evidence of sanitization or filtering was found for the text labels retrieved from diagrams before they are processed by the agent.

drawio