The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references Google Fonts (fonts.googleapis.com) to provide professional typography for the generated diagrams. It also utilizes the @resvg/resvg-js package from the official npm registry for image rendering. These are standard resources for the skill's primary functionality.
[INDIRECT_PROMPT_INJECTION]: The skill operates by processing untrusted user input (system descriptions) to generate structured diagram schemas.
Ingestion points: System descriptions provided by the user enter the agent's context during the diagram definition phase.
Boundary markers: While explicit delimiters are not defined in the schema extraction phase, the skill uses a two-step process (JSON schema generation followed by rendering) which separates data from logic.
Capability inventory: The skill possesses file-writing capabilities (creating HTML/SVG files) and command execution via the export.sh script.
Sanitization: The skill includes a critical security instruction to HTML-entity-escape all user-provided text (labels, titles, and descriptions) before rendering to prevent Cross-Site Scripting (XSS).
[COMMAND_EXECUTION]: The provided export.sh script executes shell commands to perform format conversion (HTML to PNG/PDF). The script is well-hardened, featuring local path validation to prevent directory traversal and using mktemp with restricted permissions (chmod 600) for temporary file handling.

gm-architecture