The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the draw.io desktop application CLI and xvfb-run (on Linux) to export XML diagrams to various image formats. These commands are necessary for the skill's core functionality.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a theoretical vulnerability surface where untrusted data (user diagram requirements and feedback) is processed to generate XML files and execute CLI commands.
Ingestion points: The agent ingests user requests for diagrams and feedback during the review loop in SKILL.md (Step 6).
Boundary markers: The skill suggests using XML entities for escaping special characters in values, which acts as a partial boundary, but does not implement strict sanitization for all user-provided metadata.
Capability inventory: The agent has the capability to write .drawio XML files to the working directory and execute the draw.io CLI to generate PNG, SVG, PDF, and JPG files.
Sanitization: Instructions include escaping special characters (&, <, >, ") within XML attributes, which helps mitigate basic injection within the XML structure itself.

drawio